Feast Python API Documentation¶
Feature Store¶
-
class
feast.feature_store.
FeatureStore
(repo_path: Optional[str] = None, config: Optional[feast.repo_config.RepoConfig] = None)[source]¶ Bases:
object
A FeatureStore object is used to define, create, and retrieve features.
- Parameters
repo_path – Path to a feature_store.yaml used to configure the feature store
config (RepoConfig) – Configuration object used to configure the feature store
-
apply
(objects: Union[feast.entity.Entity, feast.feature_view.FeatureView, List[Union[feast.feature_view.FeatureView, feast.entity.Entity]]])[source]¶ Register objects to metadata store and update related infrastructure.
The apply method registers one or more definitions (e.g., Entity, FeatureView) and registers or updates these objects in the Feast registry. Once the registry has been updated, the apply method will update related infrastructure (e.g., create tables in an online store) in order to reflect these new definitions. All operations are idempotent, meaning they can safely be rerun.
- Parameters
objects (List[Union[FeatureView, Entity]]) – A list of FeatureView or Entity objects that should be registered
Examples
Register a single Entity and FeatureView.
>>> from feast.feature_store import FeatureStore >>> from feast import Entity, FeatureView, Feature, ValueType, FileSource >>> from datetime import timedelta >>> >>> fs = FeatureStore() >>> customer_entity = Entity(name="customer", value_type=ValueType.INT64, description="customer entity") >>> customer_feature_view = FeatureView( >>> name="customer_fv", >>> entities=["customer"], >>> features=[Feature(name="age", dtype=ValueType.INT64)], >>> input=FileSource(path="file.parquet", event_timestamp_column="timestamp"), >>> ttl=timedelta(days=1) >>> ) >>> fs.apply([customer_entity, customer_feature_view])
-
config
: feast.repo_config.RepoConfig¶
-
delete_feature_view
(name: str)[source]¶ Deletes a feature view or raises an exception if not found.
- Parameters
name – Name of feature view
-
get_entity
(name: str) → feast.entity.Entity[source]¶ Retrieves an entity.
- Parameters
name – Name of entity
- Returns
Returns either the specified entity, or raises an exception if none is found
-
get_feature_view
(name: str) → feast.feature_view.FeatureView[source]¶ Retrieves a feature view.
- Parameters
name – Name of feature view
- Returns
Returns either the specified feature view, or raises an exception if none is found
-
get_historical_features
(entity_df: Union[pandas.core.frame.DataFrame, str], feature_refs: List[str]) → feast.infra.offline_stores.offline_store.RetrievalJob[source]¶ Enrich an entity dataframe with historical feature values for either training or batch scoring.
This method joins historical feature data from one or more feature views to an entity dataframe by using a time travel join.
Each feature view is joined to the entity dataframe using all entities configured for the respective feature view. All configured entities must be available in the entity dataframe. Therefore, the entity dataframe must contain all entities found in all feature views, but the individual feature views can have different entities.
Time travel is based on the configured TTL for each feature view. A shorter TTL will limit the amount of scanning that will be done in order to find feature data for a specific entity key. Setting a short TTL may result in null values being returned.
- Parameters
entity_df (Union[pd.DataFrame, str]) – An entity dataframe is a collection of rows containing all entity columns (e.g., customer_id, driver_id) on which features need to be joined, as well as a event_timestamp column used to ensure point-in-time correctness. Either a Pandas DataFrame can be provided or a string SQL query. The query must be of a format supported by the configured offline store (e.g., BigQuery)
feature_refs – A list of features that should be retrieved from the offline store. Feature references are of the format “feature_view:feature”, e.g., “customer_fv:daily_transactions”.
- Returns
RetrievalJob which can be used to materialize the results.
Examples
Retrieve historical features using a BigQuery SQL entity dataframe
>>> from feast.feature_store import FeatureStore >>> >>> fs = FeatureStore(config=RepoConfig(provider="gcp")) >>> retrieval_job = fs.get_historical_features( >>> entity_df="SELECT event_timestamp, order_id, customer_id from gcp_project.my_ds.customer_orders", >>> feature_refs=["customer:age", "customer:avg_orders_1d", "customer:avg_orders_7d"] >>> ) >>> feature_data = job.to_df() >>> model.fit(feature_data) # insert your modeling framework here.
-
get_online_features
(feature_refs: List[str], entity_rows: List[Dict[str, Any]]) → feast.online_response.OnlineResponse[source]¶ Retrieves the latest online feature data.
Note: This method will download the full feature registry the first time it is run. If you are using a remote registry like GCS or S3 then that may take a few seconds. The registry remains cached up to a TTL duration (which can be set to infinitey). If the cached registry is stale (more time than the TTL has passed), then a new registry will be downloaded synchronously by this method. This download may introduce latency to online feature retrieval. In order to avoid synchronous downloads, please call refresh_registry() prior to the TTL being reached. Remember it is possible to set the cache TTL to infinity (cache forever).
- Parameters
feature_refs – List of feature references that will be returned for each entity. Each feature reference should have the following format: “feature_table:feature” where “feature_table” & “feature” refer to the feature and feature table names respectively. Only the feature name is required.
entity_rows – A list of dictionaries where each key-value is an entity-name, entity-value pair.
- Returns
OnlineResponse containing the feature data in records.
Examples
>>> from feast import FeatureStore >>> >>> store = FeatureStore(repo_path="...") >>> feature_refs = ["sales:daily_transactions"] >>> entity_rows = [{"customer_id": 0},{"customer_id": 1}] >>> >>> online_response = store.get_online_features( >>> feature_refs, entity_rows) >>> online_response_dict = online_response.to_dict() >>> print(online_response_dict) {'sales:daily_transactions': [1.1,1.2], 'sales:customer_id': [0,1]}
-
list_entities
(allow_cache: bool = False) → List[feast.entity.Entity][source]¶ Retrieve a list of entities from the registry
- Parameters
allow_cache (bool) – Whether to allow returning entities from a cached registry
- Returns
List of entities
-
list_feature_views
() → List[feast.feature_view.FeatureView][source]¶ Retrieve a list of feature views from the registry
- Returns
List of feature views
-
materialize
(start_date: datetime.datetime, end_date: datetime.datetime, feature_views: Optional[List[str]] = None) → None[source]¶ Materialize data from the offline store into the online store.
This method loads feature data in the specified interval from either the specified feature views, or all feature views if none are specified, into the online store where it is available for online serving.
- Parameters
start_date (datetime) – Start date for time range of data to materialize into the online store
end_date (datetime) – End date for time range of data to materialize into the online store
feature_views (List[str]) – Optional list of feature view names. If selected, will only run materialization for the specified feature views.
Examples
Materialize all features into the online store over the interval from 3 hours ago to 10 minutes ago.
>>> from datetime import datetime, timedelta >>> from feast.feature_store import FeatureStore >>> >>> fs = FeatureStore(config=RepoConfig(provider="gcp")) >>> fs.materialize( >>> start_date=datetime.utcnow() - timedelta(hours=3), end_date=datetime.utcnow() - timedelta(minutes=10) >>> )
-
materialize_incremental
(end_date: datetime.datetime, feature_views: Optional[List[str]] = None) → None[source]¶ Materialize incremental new data from the offline store into the online store.
This method loads incremental new feature data up to the specified end time from either the specified feature views, or all feature views if none are specified, into the online store where it is available for online serving. The start time of the interval materialized is either the most recent end time of a prior materialization or (now - ttl) if no such prior materialization exists.
- Parameters
end_date (datetime) – End date for time range of data to materialize into the online store
feature_views (List[str]) – Optional list of feature view names. If selected, will only run materialization for the specified feature views.
Examples
Materialize all features into the online store up to 5 minutes ago.
>>> from datetime import datetime, timedelta >>> from feast.feature_store import FeatureStore >>> >>> fs = FeatureStore(config=RepoConfig(provider="gcp", registry="gs://my-fs/", project="my_fs_proj")) >>> fs.materialize_incremental(end_date=datetime.utcnow() - timedelta(minutes=5))
-
property
project
¶
-
refresh_registry
()[source]¶ Fetches and caches a copy of the feature registry in memory.
Explicitly calling this method allows for direct control of the state of the registry cache. Every time this method is called the complete registry state will be retrieved from the remote registry store backend (e.g., GCS, S3), and the cache timer will be reset. If refresh_registry() is run before get_online_features() is called, then get_online_feature() will use the cached registry instead of retrieving (and caching) the registry itself.
Additionally, the TTL for the registry cache can be set to infinity (by setting it to 0), which means that refresh_registry() will become the only way to update the cached registry. If the TTL is set to a value greater than 0, then once the cache becomes stale (more time than the TTL has passed), a new cache will be downloaded synchronously, which may increase latencies if the triggering method is get_online_features()
-
repo_path
: pathlib.Path¶
Config¶
-
class
feast.repo_config.
DatastoreOnlineStoreConfig
(*, type: typing_extensions.Literal[datastore] = 'datastore', project_id: pydantic.types.StrictStr = None)[source]¶ Online store config for GCP Datastore
-
project_id
: Optional[pydantic.types.StrictStr]¶ (optional) GCP Project Id
-
type
: typing_extensions.Literal[datastore]¶ Online store type selector
-
-
class
feast.repo_config.
RegistryConfig
(*, path: pydantic.types.StrictStr, cache_ttl_seconds: pydantic.types.StrictInt = 600)[source]¶ Metadata Store Configuration. Configuration that relates to reading from and writing to the Feast registry.
-
cache_ttl_seconds
: pydantic.types.StrictInt¶ The cache TTL is the amount of time registry state will be cached in memory. If this TTL is exceeded then the registry will be refreshed when any feature store method asks for access to registry state. The TTL can be set to infinity by setting TTL to 0 seconds, which means the cache will only be loaded once and will never expire. Users can manually refresh the cache by calling feature_store.refresh_registry()
- Type
-
-
class
feast.repo_config.
RepoConfig
(*, registry: Union[pydantic.types.StrictStr, feast.repo_config.RegistryConfig] = 'data/registry.db', project: pydantic.types.StrictStr, provider: pydantic.types.StrictStr, online_store: Union[feast.repo_config.DatastoreOnlineStoreConfig, feast.repo_config.SqliteOnlineStoreConfig] = SqliteOnlineStoreConfig(type='sqlite', path='data/online.db'))[source]¶ Repo config. Typically loaded from feature_store.yaml
-
online_store
: Union[feast.repo_config.DatastoreOnlineStoreConfig, feast.repo_config.SqliteOnlineStoreConfig]¶ Online store configuration (optional depending on provider)
- Type
OnlineStoreConfig
-
project
: pydantic.types.StrictStr¶ Feast project id. This can be any alphanumeric string up to 16 characters. You can have multiple independent feature repositories deployed to the same cloud provider account, as long as they have different project ids.
- Type
-
registry
: Union[pydantic.types.StrictStr, feast.repo_config.RegistryConfig]¶ Path to metadata store. Can be a local path, or remote object storage path, e.g. a GCS URI
- Type
-
-
class
feast.repo_config.
SqliteOnlineStoreConfig
(*, type: typing_extensions.Literal[sqlite] = 'sqlite', path: pydantic.types.StrictStr = 'data/online.db')[source]¶ Online store config for local (SQLite-based) store
-
path
: pydantic.types.StrictStr¶ (optional) Path to sqlite db
-
type
: typing_extensions.Literal[sqlite]¶ Online store type selector
-
Data Source¶
-
class
feast.data_source.
BigQueryOptions
(table_ref: Optional[str], query: Optional[str])[source]¶ DataSource BigQuery options used to source features from BigQuery query
-
classmethod
from_proto
(bigquery_options_proto: feast.core.DataSource_pb2.BigQueryOptions)[source]¶ Creates a BigQueryOptions from a protobuf representation of a BigQuery option
- Parameters
bigquery_options_proto – A protobuf representation of a DataSource
- Returns
Returns a BigQueryOptions object based on the bigquery_options protobuf
-
property
query
¶ Returns the BigQuery SQL query referenced by this source
-
property
table_ref
¶ Returns the table ref of this BQ table
-
classmethod
-
class
feast.data_source.
BigQuerySource
(event_timestamp_column: str, table_ref: Optional[str] = None, created_timestamp_column: Optional[str] = '', field_mapping: Optional[Dict[str, str]] = None, date_partition_column: Optional[str] = '', query: Optional[str] = None)[source]¶ -
property
bigquery_options
¶ Returns the bigquery options of this data source
-
property
-
class
feast.data_source.
DataSource
(event_timestamp_column: str, created_timestamp_column: Optional[str] = '', field_mapping: Optional[Dict[str, str]] = None, date_partition_column: Optional[str] = '')[source]¶ DataSource that can be used source features
-
property
created_timestamp_column
¶ Returns the created timestamp column of this data source
-
property
date_partition_column
¶ Returns the date partition column of this data source
-
property
event_timestamp_column
¶ Returns the event timestamp column of this data source
-
property
field_mapping
¶ Returns the field mapping of this data source
-
property
-
class
feast.data_source.
FileOptions
(file_format: Optional[feast.data_format.FileFormat], file_url: Optional[str])[source]¶ DataSource File options used to source features from a file
-
property
file_format
¶ Returns the file format of this file
-
property
file_url
¶ Returns the file url of this file
-
classmethod
from_proto
(file_options_proto: feast.core.DataSource_pb2.FileOptions)[source]¶ Creates a FileOptions from a protobuf representation of a file option
- Parameters
file_options_proto – a protobuf representation of a datasource
- Returns
Returns a FileOptions object based on the file_options protobuf
-
property
-
class
feast.data_source.
FileSource
(event_timestamp_column: str, file_url: Optional[str] = None, path: Optional[str] = None, file_format: Optional[feast.data_format.FileFormat] = None, created_timestamp_column: Optional[str] = '', field_mapping: Optional[Dict[str, str]] = None, date_partition_column: Optional[str] = '')[source]¶ -
property
file_options
¶ Returns the file options of this data source
-
property
path
¶ Returns the file path of this feature data source
-
property
-
class
feast.data_source.
KafkaOptions
(bootstrap_servers: str, message_format: feast.data_format.StreamFormat, topic: str)[source]¶ DataSource Kafka options used to source features from Kafka messages
-
property
bootstrap_servers
¶ Returns a comma-separated list of Kafka bootstrap servers
-
classmethod
from_proto
(kafka_options_proto: feast.core.DataSource_pb2.KafkaOptions)[source]¶ Creates a KafkaOptions from a protobuf representation of a kafka option
- Parameters
kafka_options_proto – A protobuf representation of a DataSource
- Returns
Returns a BigQueryOptions object based on the kafka_options protobuf
-
property
message_format
¶ Returns the data format that is used to encode the feature data in Kafka messages
-
to_proto
() → feast.core.DataSource_pb2.KafkaOptions[source]¶ Converts an KafkaOptionsProto object to its protobuf representation.
- Returns
KafkaOptionsProto protobuf
-
property
topic
¶ Returns the Kafka topic to collect feature data from
-
property
-
class
feast.data_source.
KafkaSource
(event_timestamp_column: str, bootstrap_servers: str, message_format: feast.data_format.StreamFormat, topic: str, created_timestamp_column: Optional[str] = '', field_mapping: Optional[Dict[str, str]] = {}, date_partition_column: Optional[str] = '')[source]¶ -
property
kafka_options
¶ Returns the kafka options of this data source
-
property
-
class
feast.data_source.
KinesisOptions
(record_format: feast.data_format.StreamFormat, region: str, stream_name: str)[source]¶ DataSource Kinesis options used to source features from Kinesis records
-
classmethod
from_proto
(kinesis_options_proto: feast.core.DataSource_pb2.KinesisOptions)[source]¶ Creates a KinesisOptions from a protobuf representation of a kinesis option
- Parameters
kinesis_options_proto – A protobuf representation of a DataSource
- Returns
Returns a KinesisOptions object based on the kinesis_options protobuf
-
property
record_format
¶ Returns the data format used to encode the feature data in the Kinesis records.
-
property
region
¶ Returns the AWS region of Kinesis stream
-
property
stream_name
¶ Returns the Kinesis stream name to obtain feature data from
-
classmethod
-
class
feast.data_source.
KinesisSource
(event_timestamp_column: str, created_timestamp_column: str, record_format: feast.data_format.StreamFormat, region: str, stream_name: str, field_mapping: Optional[Dict[str, str]] = {}, date_partition_column: Optional[str] = '')[source]¶ -
property
kinesis_options
¶ Returns the kinesis options of this data source
-
property
Entity¶
-
class
feast.entity.
Entity
(name: str, value_type: feast.value_type.ValueType, description: str = '', join_key: Optional[str] = None, labels: Optional[MutableMapping[str, str]] = None)[source]¶ Represents a collection of entities and associated metadata.
-
property
created_timestamp
¶ Returns the created_timestamp of this entity
-
property
description
¶ Returns the description of this entity
-
classmethod
from_dict
(entity_dict)[source]¶ Creates an entity from a dict
- Parameters
entity_dict – A dict representation of an entity
- Returns
Returns a EntityV2 object based on the entity dict
-
classmethod
from_proto
(entity_proto: feast.core.Entity_pb2.Entity)[source]¶ Creates an entity from a protobuf representation of an entity
- Parameters
entity_proto – A protobuf representation of an entity
- Returns
Returns a EntityV2 object based on the entity protobuf
-
classmethod
from_yaml
(yml: str)[source]¶ Creates an entity from a YAML string body or a file path
- Parameters
yml – Either a file path containing a yaml file or a YAML string
- Returns
Returns a EntityV2 object based on the YAML file
-
is_valid
()[source]¶ Validates the state of a entity locally. Raises an exception if entity is invalid.
-
property
join_key
¶ Returns the join key of this entity
-
property
labels
¶ Returns the labels of this entity. This is the user defined metadata defined as a dictionary.
-
property
last_updated_timestamp
¶ Returns the last_updated_timestamp of this entity
-
property
name
¶ Returns the name of this entity
-
to_dict
() → Dict[source]¶ Converts entity to dict
- Returns
Dictionary object representation of entity
-
to_proto
() → feast.core.Entity_pb2.Entity[source]¶ Converts an entity object to its protobuf representation
- Returns
EntityV2Proto protobuf
-
to_spec_proto
() → feast.core.Entity_pb2.EntitySpecV2[source]¶ Converts an EntityV2 object to its protobuf representation. Used when passing EntitySpecV2 object to Feast request.
- Returns
EntitySpecV2 protobuf
-
to_yaml
()[source]¶ Converts a entity to a YAML string.
- Returns
Entity string returned in YAML format
-
property
value_type
¶ Returns the type of this entity
-
property
Feature View¶
-
class
feast.feature_view.
FeatureView
(name: str, entities: List[str], features: List[feast.feature.Feature], ttl: Optional[Union[google.protobuf.duration_pb2.Duration, datetime.timedelta]], input: Union[feast.data_source.BigQuerySource, feast.data_source.FileSource], tags: Optional[Dict[str, str]] = None, online: bool = True)[source]¶ A FeatureView defines a logical grouping of serveable features.
-
classmethod
from_proto
(feature_view_proto: feast.core.FeatureView_pb2.FeatureView)[source]¶ Creates a feature view from a protobuf representation of a feature view
- Parameters
feature_view_proto – A protobuf representation of a feature view
- Returns
Returns a FeatureViewProto object based on the feature view protobuf
-
classmethod
Feature Table¶
-
class
feast.feature_table.
FeatureTable
(name: str, entities: List[str], features: List[feast.feature.Feature], batch_source: Optional[Union[feast.data_source.BigQuerySource, feast.data_source.FileSource]] = None, stream_source: Optional[Union[feast.data_source.KafkaSource, feast.data_source.KinesisSource]] = None, max_age: Optional[google.protobuf.duration_pb2.Duration] = None, labels: Optional[MutableMapping[str, str]] = None)[source]¶ Represents a collection of features and associated metadata.
-
add_feature
(feature: feast.feature.Feature)[source]¶ Adds a new feature to the feature table.
-
property
batch_source
¶ Returns the batch source of this feature table
-
property
created_timestamp
¶ Returns the created_timestamp of this feature table
-
property
entities
¶ Returns the entities of this feature table
-
property
features
¶ Returns the features of this feature table
-
classmethod
from_dict
(ft_dict)[source]¶ Creates a feature table from a dict
- Parameters
ft_dict – A dict representation of a feature table
- Returns
Returns a FeatureTable object based on the feature table dict
-
classmethod
from_proto
(feature_table_proto: feast.core.FeatureTable_pb2.FeatureTable)[source]¶ Creates a feature table from a protobuf representation of a feature table
- Parameters
feature_table_proto – A protobuf representation of a feature table
- Returns
Returns a FeatureTableProto object based on the feature table protobuf
-
classmethod
from_yaml
(yml: str)[source]¶ Creates a feature table from a YAML string body or a file path
- Parameters
yml – Either a file path containing a yaml file or a YAML string
- Returns
Returns a FeatureTable object based on the YAML file
-
is_valid
()[source]¶ Validates the state of a feature table locally. Raises an exception if feature table is invalid.
-
property
labels
¶ Returns the labels of this feature table. This is the user defined metadata defined as a dictionary.
-
property
last_updated_timestamp
¶ Returns the last_updated_timestamp of this feature table
-
property
max_age
¶ Returns the maximum age of this feature table. This is the total maximum amount of staleness that will be allowed during feature retrieval for each specific feature that is looked up.
-
property
name
¶ Returns the name of this feature table
-
property
stream_source
¶ Returns the stream source of this feature table
-
to_dict
() → Dict[source]¶ Converts feature table to dict
- Returns
Dictionary object representation of feature table
-
to_proto
() → feast.core.FeatureTable_pb2.FeatureTable[source]¶ Converts an feature table object to its protobuf representation
- Returns
FeatureTableProto protobuf
-
Feature¶
-
class
feast.feature.
Feature
(name: str, dtype: feast.value_type.ValueType, labels: Optional[MutableMapping[str, str]] = None)[source]¶ Feature field type
-
property
dtype
¶ Getter for data type of this field
-
classmethod
from_proto
(feature_proto: feast.core.Feature_pb2.FeatureSpecV2)[source]¶ - Parameters
feature_proto – FeatureSpecV2 protobuf object
- Returns
Feature object
-
property
labels
¶ Getter for labels of this field
-
property
name
¶ Getter for name of this field
-
property
-
class
feast.feature.
FeatureRef
(name: str, feature_table: str)[source]¶ Feature Reference represents a reference to a specific feature.
-
classmethod
from_proto
(proto: feast.serving.ServingService_pb2.FeatureReferenceV2)[source]¶ Construct a feature reference from the given FeatureReference proto
- Parameters
proto – Protobuf FeatureReference to construct from
- Returns
FeatureRef that refers to the given feature
-
classmethod
from_str
(feature_ref_str: str)[source]¶ Parse the given string feature reference into FeatureRef model String feature reference should be in the format feature_table:feature. Where “feature_table” and “name” are the feature_table name and feature name respectively.
- Parameters
feature_ref_str – String representation of the feature reference
- Returns
FeatureRef that refers to the given feature
-
classmethod