**Build a feature store with point-in-time correct joins, offline/online stores, and a feature registry.**
Browse the actual product documentation and code examples included in this toolkit.
Key features of Feature Store Setup
• **Feature registry** for defining, versioning, searching, and validating features with ownership metadata • **Offline store** (local Parquet reference impl) for batch feature retrieval during training • **Online store** (in-memory reference impl) for low-latency feature serving at inference time • **Point-in-time correct joins** that prevent training/serving skew and data leakage • **Feature versioning** with change tracking, diffs, and rollback support • **Materialization scheduler** for computing features on configurable cadences
**Feature registry** for defining, versioning, searching, and validating features with ownership metadata
**Offline store** (local Parquet reference impl) for batch feature retrieval during training
**Online store** (in-memory reference impl) for low-latency feature serving at inference time
**Point-in-time correct joins** that prevent training/serving skew and data leakage
**Feature versioning** with change tracking, diffs, and rollback support
**Materialization scheduler** for computing features on configurable cadences
Configure Feature Store Setup parameters to see how the product works.
pip install -r requirements.txt
python -c "
from src.feature_store import FeatureStore
import pandas as pd
store = FeatureStore()
df = pd.DataFrame({
'customer_id': [1, 2, 3],
'avg_spend': [100.0, 200.0, 150.0],
'event_time': pd.Timestamp.now(),
})
store.materialize('avg_spend', df, 'cu