Package bytehub

ByteHub provides an easy-to-use Feature Store, optimised for time-series data. It requires no complex infrastructure setup, and can be run locally, connected to a remote database/file storage, or run in a cloud-hosted mode. ByteHub uses Dask for data storage, allowing it to be scaled to large datasets and cluster compute environments.

Example usage for a local SQLite feature store:

import bytehub as bh
fs = bh.FeatureStore()

Remote feature stores can be accessed using a SQLAlchemy connection string, e.g.:

fs = bh.FeatureStore('postgresql+psycopg2://user:pass@host:port/bytehub')

Cloud-hosted feature stores can be accessed via a REST API endpoint:

fs = bh.FeatureStore('')

or simply use the following to access ByteHub's cloud service:

fs = bh.CloudFeatureStore()

See for examples and tutorials.


Exception raised by Feature Store objects


def FeatureStore(connection_string='sqlite:///bytehub.db', **kwargs)

Factory method to create Feature Store objects.


connection_string : str
SQLAlchemy connection string for database containing feature store metadata (defaults to local sqlite file) or an HTTPS endpoint to a cloud-hosted feature store.
Additional options to be passed to the Feature Store constructor.


Union[CoreFeatureStore, CloudFeatureStore]
Feature Store object.