How to manage data in Feast feature stores
Feast (Feature Store) is an operational data system for managing and serving machine learning features to models in production. Feast is able to serve feature data to models from a low-latency online store (for real-time prediction) or from an offline store (for scale-out batch scoring or model training).

When would you want to use it?

There are two core functions that feature stores enable:
  • access to data from an offline / batch store for training.
  • access to online data at inference time.
Feast integration currently supports your choice of offline data sources, and a Redis backend for your online feature serving. We encourage users to check out Feast's documentation and guides on how to set up your offline and online data sources via the configuration yaml file.
COMING SOON: While the ZenML integration has an interface to access online feature store data, it currently is not usable in production settings with deployed models. We will update the docs when we enable this functionality.

How to deploy it?

The Feast Feature Store flavor is provided by the Feast ZenML integration, you need to install it, to be able to register it as a Feature Store and add it to your stack:
zenml integration install feast
Since this example is built around a Redis use case, a Python package to interact with Redis will get installed alongside Feast, but you will still first need to install Redis yourself. See this page for some instructions on how to do that on your operating system.
You will then need to run a Redis server in the background in order for this example to work. You can either use the redis-server command in your terminal (which will run a continuous process until you CTRL-C out of it), or you can run the daemonized version:
redis-server --daemonize yes
# verify it is running (Unix machines)
ps aux | grep redis-server

How do you use it?

ZenML assumes that users already have a feature store that they just need to connect with. The ZenML Online data retrieval is currently possible in a local setting, but we don't currently support using the online data serving in the context of a deployed model or as part of model deployment. We will update this documentation as we develop out this feature.
ZenML supports access to your feature store via a stack component that you can configure via the CLI tool. ( See here for details on how to do that.)
Getting features from a registered and active feature store is possible by creating your own step that interfaces into the feature store:
from datetime import datetime
from typing import Any, Dict, List, Union
import pandas as pd
from zenml.steps import BaseStepConfig, step, StepContext
entity_dict = {} # defined in earlier code
features = [] # defined in earlier code
class FeastHistoricalFeaturesConfig(BaseStepConfig):
"""Feast Feature Store historical data step configuration."""
entity_dict: Union[Dict[str, Any], str]
features: List[str]
full_feature_names: bool = False
class Config:
arbitrary_types_allowed = True
def get_historical_features(
config: FeastHistoricalFeaturesConfig,
context: StepContext,
) -> pd.DataFrame:
"""Feast Feature Store historical data step
config: The step configuration.
context: The step context.
The historical features as a DataFrame.
if not context.stack:
raise DoesNotExistException(
"No active stack is available. Please make sure that you have registered and set a stack."
elif not context.stack.feature_store:
raise DoesNotExistException(
"The Feast feature store component is not available. "
"Please make sure that the Feast stack component is registered as part of your current active stack."
feature_store_component = context.stack.feature_store
config.entity_dict["event_timestamp"] = [
for val in config.entity_dict["event_timestamp"]
entity_df = pd.DataFrame.from_dict(config.entity_dict)
return feature_store_component.get_historical_features(
historical_features = get_historical_features(
entity_dict=historical_entity_dict, features=features
Note that ZenML's use of Pydantic to serialize and deserialize inputs stored in the ZenML metadata means that we are limited to basic data types. Pydantic cannot handle Pandas DataFrames, for example, or datetime values, so in the above code you can see that we have to convert them at various points.
A concrete example of using the Feast feature store can be found here.
For more information and a full list of configurable attributes of the Kubeflow orchestrator, check out the API Docs.