Building your data model
Entities are the main concept inside Graft with which you will be interacting. Entities will contain fields with your unstructured data. Graft will transform selected data fields by creating representations known as embeddings. Once your entity exists and your data has been transformed, you can optionally augment it by applying Enrichments which predict properties about your data.
Entities are specific to a project, it is possible to nest entities within a project to support different use cases. Entities are also the underlying data structure used by Graft Apps.
All this information, data fields, embeddings, enrichments, are held within the entity and available via SQL and the Graft API
Entities can be found under the project navigation menu
Entity creation step-by-step
From the Entity navigation tab Click on the CREATE ENTITY button, this starts a workflow to step you through the process of creating a new Entity
- Name your new entity
- Add a suitable description so other users can understand the entity’s purpose
- Click NEXT
The next step is to select the fields you wish to pull into this entity from the data sources associated with your project. Initially you will be presented with ALL of the data sources and their fields. Select the data source you would like to work with from the drop down menu to only show the fields associated with it.
FIELDS FROM MULTIPLE SOURCES
You can select fields for your entity from one or more data sources depending on your needs
- Check each of the fields you need including an index/key field (in this example the id field)
- Rename any field if required
- Click NEXT
Primary Key Selection
Graft needs to know which of your selected fields should be the Primary Key (this is important when linking to other entities and for labeling)
- Click the appropriate field
- Click NEXT
Define Embeddings (optional)
SELECTING THE BEST FIELD FOR EMBEDDINGS
Most often, you will want to select the field containing unstructured data like text or images. This is also the field that will be transformed by Graft using Machine Learning models so you can do things like similarity search and enrichments downstream
To create embeddings for any of the fields in the entity Check each field of interest AND also select from the available Trunk Models using the drop down menu. The default model will be displayed for the field type (image, text etc.)
Graft supports multiple options for Embedding configuration within an Entity. For further information please see Entity embedding options.
Advanced settings allows the user to change the embedding strategy for a given trunk. We recommend this for advanced user only and is done in conjunction with reading our Embedding Strategy documentation
Once all selections have been made Click FINISH, you will be returned to the Entity Dashboard Your Entity is now configured, but not processed.
At this time you can select one of the preconfigured processing workflows