Chatting With The Data Analyst

Prerequisite

Before chatting with our AI Data Analyst, make sure you have created/uploaded/connected a Data Source. For your data sources ensure the data schema is in english as Turbular is not designed to work with non english data schemas and performance may be heavily degraded. For Api data sources this does not apply. While the schema needs to be in english, individual data points can be in whatever language you choose. For example, you have uploaded a Excel File about sales data in german, naming the columns of the file "Order Date","Order Value","City" is fine, while using german columns like "Bestelldatum", "Verkaufswert", "Stadt" would lead to degraded results.

More information about how to set up a data source can be found at Setting up Datasources.

Getting Started

After entering the Chat you are presented with a list of type of data source (Database, Excel, or Api). After selecting a type an overview of available data sources of the selected type is presented to you. Note that only for Apis selecting multiple data sources is supported. An overview of the available agents can be found at Available Chat Agents. Once an available data source and agent is selected the list of data sources disappears and messages can be sent to the agent.

Capabilities Of The Agent

Our AI Data Analyst agents are able to:

• Connect to your data source
• Retrieve data from your data source
• Visualize data
• Respond to your message

Note while Turbular is designed to be a data analytics platform there is nothing stopping you from engaging in a casual chat with our agents. More information about general capabilities of Turbular can be found at Capabilities of our AI Agents.

Chatting

After sending a message to the agent, the agent will start to process your request. Depending on your message the agent will decide whether to retrieve data from your data source. If the agent decided to retrieve data it will then decide whether to visualize the retrieved data. An overview of the visualizations the agent has access to is provided at Supported Visualizations. Finally, the agent will respond to your message with text (in cases where the agent generated a visualization it may not generate a text response).

Interacting with Agent Replies

Inspecting Agent Response

After receiving a reply from our agent it is possible to further inspect the message generated by the agent via the "Message Details" button. After pressing the button a new pop up window will open over the message giving you access to two tabs. In the Chart tab you can edit the chart generated by the AI Data Analyst if it generated a chart. In the Query tab you are presented with the data that has been retrieved and the SQL statement used on your data source to retrieve the data. Please note that we use SQL as well in order to represent requests against Apis.

Both the chart as well as the retrieved data can be downloaded via this Export button.

Rating Agent Response

Using the the thumbs up and down button you can further give feedback about the quality of their answers to our agents. This will lead to them being better able to serve your requests in the future.

Information Access for Agent Response

At the bottom of each message a tag or multiple tags (in the case of an API data source) can be present. This is in order to inform you about which data source where used in order to generate the associated reply.

Limitations

Agent Limitations

Our AI Data Analyst excels at retrieving information. Therefore, all question related to any form of data retrival can be answered with no problem. Limitations arise for questions that are targeted towards inference (f.e "Why is most of my sales revenue in New York"), as while AI seem smart this capability seems to be luckily still limited to human intelligence.

Data Limitations

Our agents can only operate on your data and are therefore limited by the quality of your data. If you provide them with a badly labeled dataset with bad wording our agents will struggle to make sense of your data.

Subscription Limitation

For free users Turbular only supports retrieving 6000 data rows from a data source per request. Additionally agent capabilities and response quality varies based on the selected agent.

Example

Supported Languages

You can chat with Turbular's AI Data Analysts in the following languages:

• English
• French
• German
• Spanish
• Dutch
• Italian
• Polish
• Afrikaans
• Japanese
• Chinese
• Indonesian
• Arabic
• Greek
• Turkish