Table of contents
Official Content
  • This documentation is valid for:

This document is intended to provide guidelines on how to ask questions to get information that was ingested in a particular Chat with Data Assistant.

General Considerations

Use Natural Language

Type your question naturally just like you would ask a colleague for information.

It doesn't matter if you make spelling mistakes or use different synonyms.

It is important to convey your intention to the assistant correctly.

For example, suppose you have a database with information about continents/countries and want to get the list of continents together with the number of countries in each one. In this case, your query can be formulated in one of the following ways:

  • Number of countries per continent.
  • How many countries are there in each continent?
  • Give me the list of continents with the number of countries in each continent.

Iterate and refine your questions

It is not necessary to repeat all the previous information, as this is inferred from the conversation.

For example, if you previously asked "How many countries are there in each continent?", you can follow up with "Only those with more than 50 countries". In this way, the assistant will understand that you want the list of continents and the number of countries for each one, but limited to those with more than 50 countries in total.

Remember that with each new question, you can add more details or change the focus as needed.

Questions that the Assistant can answer

The assistant is designed to provide answers from the database and also from the database metadata.

When a question is asked, the assistant first determines whether it is necessary to access the database to return the answer. If the question is about metadata (and therefore it isn't necessary to access the database to look for data), it can be answered using natural language. On the other hand, if the question requires access to the database, the assistant builds a Query and the possible Output Formats are Table, PivotTable, Chart or Card (no free text).

Examples of metadata queries

  • What can I ask about?
  • What entities are available?
  • Can you suggest some questions I could ask?

Examples of data queries

  • Show the top 10 countries by GDP per capita.
  • Number of patients by continent.
  • List the top 10 countries with the highest population.

Questions that the Assistant can't answer

There are certain types of queries that the assistant will not be able to answer, or whose answer may not be as expected. 

Information not contained in the database

If the requested information is not in the database, the assistant will not be able to provide an answer.

For example, if the database is only about continents/countries:

Question: Total exports in the last year.
Answer: The answer is not found in the entities provided because there is no entity or attribute related to export data or economic transactions.

Concepts not represented in the database

The assistant does not have information about the concepts handled in the organization but not included in the database.

Do: Number of students promoted on the 05/01/2024 exam.
Don't: Number of students with a grade higher than 6 on the 05/01/2024 exam.

Calculations or Formulas

It is not possible to request calculations based on known formulas referenced by name. For example, requesting the Human Development Index for a country will not work, even if the elements to calculate it are in the database.

Predictions

The assistant can only provide information based on the current data contained in the database. It cannot make predictions unless they are specifically recorded in a database table.

 

Last update: March 2025 | © GeneXus. All rights reserved. GeneXus Powered by Globant