Skip to main content

Features

View as Markdown

Create and manage experiments, features, and other machine learning artifacts

Batch create materialized features Beta​

POST /api/2.0/feature-engineering/materialized-features:batchCreate

Batch create materialized features.

API scopes: mlflow

Request body​

requestsarray of objectRequired

The requests to create materialized features.

Show child attributesHide child attributes
materialized_featureobjectRequired

The materialized feature to create.

Show child attributesHide child attributes
materialized_feature_idstringIDImmutable

Server-assigned unique identifier for the materialized feature.

feature_namestringRequired

The full name of the feature in Unity Catalog.

offline_store_configobjectInput onlyImmutable

Destination for writing feature values to an offline Delta table. The resulting table is returned as table_name.

Show child attributesHide child attributes
catalog_namestringRequired

The Unity Catalog catalog name.

schema_namestringRequired

The Unity Catalog schema name.

table_name_prefixstringRequired

Prefix for Unity Catalog table name. The materialized feature will be stored in a table with this prefix and a generated postfix.

online_store_configobjectImmutable

Destination for writing feature values to an online Lakebase table. The resulting table is returned as table_name.

Show child attributesHide child attributes
catalog_namestringRequiredInput only

The Unity Catalog catalog name. This name is also used as the Lakebase logical database name. Quoting is handled by the backend where needed, do not pre-quote it.

schema_namestringRequiredInput only

The Unity Catalog schema name. This name is also used as the Lakebase schema name under the database. Quoting is handled by the backend where needed, do not pre-quote it.

table_name_prefixstringRequiredInput only

Prefix for Unity Catalog table name. The materialized feature will be stored in a Lakebase table with this prefix and a generated postfix.

online_store_namestringRequired

The name of the target online store.

cron_schedule_triggerobject

A cron-based schedule trigger for the materialization pipeline.

Show child attributesHide child attributes
cron_expressionstring

The cron expression defining the schedule (e.g., "0 0 * * *" for daily at midnight). The schedule is interpreted in timezone_id (defaults to UTC). Required when mode is MANUAL (or unset). Left empty when mode is DERIVED, where the service computes it (aligned to UTC) from the features' window timing and fills it in on the response.

table_triggerobject

A trigger that fires when the upstream source table changes.

streaming_modeobject

The Structured Streaming trigger mode used for materialization. Real-time mode (RTM) targets sub-second latency for operational workloads; micro-batch mode (MBM) favors cost efficiency for ETL and analytics workloads.

Show child attributesHide child attributes
modestring

The type of streaming mode used by the materialization pipeline.

Values:

  • STREAMING_MODE_TYPE_UNSPECIFIED
  • STREAMING_MODE_TYPE_RTM
  • STREAMING_MODE_TYPE_MBM

Response​

materialized_featuresarray of object

The created materialized features with assigned IDs.

Show child attributesHide child attributes
materialized_feature_idstringIDImmutable

Server-assigned unique identifier for the materialized feature.

feature_namestringRequired

The full name of the feature in Unity Catalog.

online_store_configobjectImmutable

Destination for writing feature values to an online Lakebase table. The resulting table is returned as table_name.

Show child attributesHide child attributes
online_store_namestringRequired

The name of the target online store.

table_namestringOutput only

The fully qualified Unity Catalog path to the table containing the materialized feature (Delta table or Lakebase table). Output only.

last_materialization_timestringOutput only

The timestamp when the pipeline last ran and updated the materialized feature values. If the pipeline has not run yet, this field will be null.

is_onlinebooleanOutput only

True if this is an online materialized feature. False if it is an offline materialized feature.

cron_schedule_triggerobject

A cron-based schedule trigger for the materialization pipeline.

Show child attributesHide child attributes
cron_expressionstring

The cron expression defining the schedule (e.g., "0 0 * * *" for daily at midnight). The schedule is interpreted in timezone_id (defaults to UTC). Required when mode is MANUAL (or unset). Left empty when mode is DERIVED, where the service computes it (aligned to UTC) from the features' window timing and fills it in on the response.

table_triggerobject

A trigger that fires when the upstream source table changes.

streaming_modeobject

The Structured Streaming trigger mode used for materialization. Real-time mode (RTM) targets sub-second latency for operational workloads; micro-batch mode (MBM) favors cost efficiency for ETL and analytics workloads.

Show child attributesHide child attributes
modestring

The type of streaming mode used by the materialization pipeline.

Values:

  • STREAMING_MODE_TYPE_UNSPECIFIED
  • STREAMING_MODE_TYPE_RTM
  • STREAMING_MODE_TYPE_MBM