API Reference¶

Timefence's Python API is organized around five core objects and four functions.

Core Objects¶

Class	Purpose
`Source`	A historical data source (Parquet, CSV, DataFrame)
`Feature`	A named signal derived from a Source
`Labels`	Prediction target definition
`FeatureSet`	Group features for reuse
`Store`	Build tracking and caching

Functions¶

Function	Purpose
`build()`	Construct a point-in-time correct training dataset
`audit()`	Scan a dataset for temporal leakage
`explain()`	Preview join logic without executing
`diff()`	Compare two datasets for changes

Quick example¶

import timefence

source = timefence.Source(path="data/users.parquet", keys=["user_id"], timestamp="updated_at")
feature = timefence.Feature(source=source, columns=["country"])
labels = timefence.Labels(path="data/labels.parquet", keys=["user_id"], label_time="label_time", target="churned")

result = timefence.build(labels=labels, features=[feature], output="train.parquet")