Feature Views are now generally available for batch and streaming

Feature Views enable you to define and compute features from sources such as Delta tables, Kafka streams, and request-time data, integrating with Unity Catalog for model training and serving.

Use cases

Imagine you need to feed machine learning models with consistent temporal aggregations and real-time data.

• Define feature objects directly in Unity Catalog.
• Compute point-in-time features for model training and inference.
• Materialize features to offline and online stores.

How to use

Install the library and define features using Python.

from databricks.feature_engineering import FeatureEngineeringClient
from databricks.feature_engineering.entities import (
    DeltaTableSource, Feature, AggregationFunction, Avg, TumblingWindow
)
from datetime import timedelta

source = DeltaTableSource(catalog_name="main", schema_name="feature_store", table_name="transactions")

avg_feature = Feature(
    source=source,
    entity=["user_id"],
    timeseries_column="transaction_time",
    function=AggregationFunction(Avg(input="amount"), TumblingWindow(window_duration=timedelta(days=30))),
    name="avg_transaction_30d",
)

fe = FeatureEngineeringClient()
feature_df = fe.compute_features(features=[avg_feature])
feature_df.display()

What happens in this code:

1. Configures the data source pointing to a Delta table.
2. Defines a rolling average feature with a 30-day window grouped by user.
3. Initializes the feature engineering client and computes the resulting DataFrame.

Limitations and requirements

• Requires serverless compute or a classic compute cluster running Databricks Runtime 17.0 ML or above.
• You must install the databricks-feature-engineering>=0.16.0 package.
• Streaming features require an online_config and do not support offline_config directly.

Using it in a pipeline

They fit right after ingestion in the Bronze/Silver layers to unify feature engineering logic between offline model training and low-latency online model serving.