This repository provides concrete code examples demonstrating how to use Tecton for feature engineering in machine learning applications. You'll find a comprehensive collection of examples implemented in two ways:
- Using Spark for Batch and Streaming features
- Using Rift, Tecton's Python-native compute engine, for Batch, Streaming and Real-Time features
Please navigate to the spark or rift folder depending on which compute you use. If you already know, use, and love Spark or Databricks, you can start with the spark folder. If you prefer using Python and/or SQL without the need for a Spark cluster, check out the rift folder.
The examples cover a wide range of feature engineering patterns and techniques, from simple aggregations to complex feature pipelines, all in the context of real-world use cases like fraud detection, recommendation systems and ads serving.
- Install tecton using
pip install tecton>=1.1 - Log into your tecton cluster using
tecton login <cluster_name> - Create a new workspace using
tecton workspace create <workspace_name> cdinto thesparkorriftfolder depending on which compute you use- Initialize repo using
tecton init - Apply the features defined here using
tecton apply
- Tecton's documentation has information about how to use Tecton to create, manage, and serve operational ML features
- CLI installation instructions