What is Qluent?

Qluent is a smart business analytics assistant, designed to enable non-technical users to access, explore and gain insights from their business data. It leverages the power of large language models (LLM), which are powering services such as ChatGPT, enabling users to interact with their data in a fully natural language environment, without any additional learning.

What sets it apart from other LLM-based services is that it is connected to your business data sources, and uses that live data when answering questions, instead of being pre-trained on a static dataset.

How does Qluent work?

All data sources, such as SQL databases or data providing APIs, define a language for querying them. Qluent is trained to transform natural language questions into queries for these data sources, and then additionally trained to query your business data source in particular, with the awareness of the data objects (e.g. SQL tables, columns, or API query fields) this data source holds. When users ask a question, Qluent first translates it into a query for the data source, and then executes it on the data source, returning the results to the user.

Additionally, Qluent provides users with an explanation of how the question was understood and how the query that was generated actually works, allowing users to understand how the system works interpreted their question and letting them guide the answering process with precision.

What models does Qluent use?

Qluent is powered by a variety of large language models, such as GPT, Flan-T5 and Bert. These models are hosted in the US and the EU, depending on where our users are located.

Does this mean that Qluent will share my data with 3rd parties?

No, unless you explicitly allow it, Qluent does not send any of your data to the underlying 3rd party models. Qluent answers questions not by directly generating answers from your data, but instead by generating code that answers the question.

This means that Qluent does not need to store your data to answer questions. Instead, it only stores the structure of your data (e.g. SQL schema) to generate the code. This code is then executed on your data infrastructure, and the results are displayed to the user. Optionally, the results will be saved by Qluent for easier access in future.

When setting up the project, you will be in full control over which information about your data structure is actually stored and shared with 3rd parties, and you can always revoke access to particular data objects at any time.

How can I tell if the answers I get are correct?

Qluent offers a couple of ways to verify the correctness of the answers it provides. First, it provides explanations of how the question was understood, letting users check, and if necessary correct, the interpretation of each part of their question. Second, it provides explanations of the code itself. Finally, Qluent has additional mechanisms in place which automatically detect if data needed to answer the question is insufficient, or if the code that was generated is incorrect.

With that said, Qluent is still a young technology, and it is possible that it will make mistakes. If you encounter a situation where you are not sure if the answer is correct, please let us know, and we will investigate the issue.