API Components

Chalk’s feature pipeline helps you define the features you need and compute their values. The following concepts are the foundation of our implementation:

• Features - Chalk lets you spell out your feature values using Python classes. Feature classes are namespaced, and can have relationships to other feature namespaces.
• Resolvers - Resolvers compute the values of your features from other feature values. Feature resolvers are defined online and offline for inference and training, and support ETL between online and offline.
• Online Inference - In production, you can query your features through one of Chalk’s API clients. At query time, you can choose to override any feature caching.
• Training - Chalk supports querying your offline features directly from a Python notebook.
• Data Export - Data can be exported from your offline feature store to any destination of your choice.