ThinkData Works integrates with large language models (LLM) in order to automate the process of metadata creation and curation. If you are using are an enterprise user, you will have selected the LLM the platform uses during the deployment phase.
For all other users, the AI co-pilot is preconfigured.
The co-pilot is designed to speed up the process of adding metadata to your data catalog by automating the addition of data descriptions, classifications, data dictionary definitions, and business glossary terms.
To use the tool, go to any dataset in your catalog that you have edit access on (you can tell you have edit access on a dataset when the "edit" icon in the dataset action sidebar is clickable).
Select "edit"
At the top of the dataset edit page you will see a button with a wand icon that looks like this:
You will also see the wand icon at the right-hand side of the dataset description editor, the classification editor, and every dataset property in the data dictionary.
By selecting any of these individual fields, the platform will automatically create and send a prompt to the preconfigured LLM and return a description, classification, or data dictionary definition and business glossary term for a specific field in the dataset.
To generate all metadata at once, select the "Generate metadata" button at the top of the dataset edit page. After a few moments, metadata suggestions will appear.
These suggestions are based on the metadata already available in the dataset, including its title and properties. If you would like to refine the prompt, you may provide additional metadata prior to selecting the co-pilot to improve the quality of the response. Additionally, the metadata fields that are provided are suggestions - you may revert, deny, accept, or alter them as needed.