16. Manage Feature Life Cycle

Manage Feature Life Cycle¶

In this tutorial, you will learn how to:

• Update Default Feature Job Settings at the table level when there are changes in source table availability or freshness.
• Modify Default Cleaning Operations at the table level when data quality issues are detected.
• Create new version of features that show discrepancies with table defaults.
• Create a new version of a feature list.
• Delete/Deprecate features, feature lists, and their associated guardrails.

---

Important Note for FeatureByte Enterprise Users with Approval Flow¶

For catalogs with Approval Flow enabled, any changes to table metadata trigger a review process.
This process automatically recommends new versions of features and feature lists linked to the affected tables, ensuring that future models and deployments use updated versions that address data integrity or freshness issues.

For a deeper dive into FeatureByte Enterprise, see:

• Deploy and Serve
• Manage Feature Life Cycle

In [93]:

import featurebyte as fb
import pandas as pd

# Set your profile to the tutorial environment
fb.use_profile("tutorial")

catalog_name = "Credit Default Dataset SDK Tutorial"
catalog = fb.Catalog.activate(catalog_name)

import featurebyte as fb import pandas as pd # Set your profile to the tutorial environment fb.use_profile("tutorial") catalog_name = "Credit Default Dataset SDK Tutorial" catalog = fb.Catalog.activate(catalog_name)

17:18:32 | INFO     | Using profile: staging
INFO    :featurebyte:Using profile: staging
17:18:32 | INFO     | Using configuration file at: /Users/gxav/.featurebyte/config.yaml
INFO    :featurebyte:Using configuration file at: /Users/gxav/.featurebyte/config.yaml
17:18:32 | INFO     | Active profile: staging (https://staging.featurebyte.com/api/v1)
INFO    :featurebyte:Active profile: staging (https://staging.featurebyte.com/api/v1)
17:18:32 | INFO     | SDK version: 3.2.0.dev66
INFO    :featurebyte:SDK version: 3.2.0.dev66
17:18:32 | INFO     | No catalog activated.
INFO    :featurebyte:No catalog activated.
17:18:32 | INFO     | Catalog activated: Credit Default Dataset SDK Tutorial
INFO    :featurebyte.api.catalog:Catalog activated: Credit Default Dataset SDK Tutorial
16:16:27 | INFO     | Active profile: tutorial (https://tutorials.featurebyte.com/api/v1)
16:16:27 | WARNING  | Remote SDK version (1.1.0.dev7) is different from local (1.1.0.dev1). Update local SDK to avoid unexpected behavior.
16:16:27 | INFO     | No catalog activated.
16:16:27 | INFO     | Catalog activated: Grocery Dataset Tutorial

Update Default Feature Job Setting at the table level¶

In [94]:

# Get BUREAU table
bureau_table = catalog.get_table("BUREAU")

# Get BUREAU table bureau_table = catalog.get_table("BUREAU")

In [95]:

# Get current Default Feature Job Setting
bureau_table.default_feature_job_setting

# Get current Default Feature Job Setting bureau_table.default_feature_job_setting

Out[95]:

FeatureJobSetting(blind_spot='25800s', period='86400s', offset='25740s', execution_buffer='0s')

In [96]:

# List past analysis
past_analysis = bureau_table.list_feature_job_setting_analysis()

# List past analysis past_analysis = bureau_table.list_feature_job_setting_analysis()

In [97]:

# Get past analysis
analysis_id = past_analysis.id.to_list()[0]
analysis = fb.FeatureJobSettingAnalysis.get_by_id(analysis_id)

# Get past analysis analysis_id = past_analysis.id.to_list()[0] analysis = fb.FeatureJobSettingAnalysis.get_by_id(analysis_id)

In [98]:

# Backtest new setting
new_feature_job_setting = fb.FeatureJobSetting(
    blind_spot='28800s',
    period=bureau_table.default_feature_job_setting.period,
    offset=bureau_table.default_feature_job_setting.offset
)
backtest_result = analysis.backtest(feature_job_setting=new_feature_job_setting)

# Backtest new setting new_feature_job_setting = fb.FeatureJobSetting( blind_spot='28800s', period=bureau_table.default_feature_job_setting.period, offset=bureau_table.default_feature_job_setting.offset ) backtest_result = analysis.backtest(feature_job_setting=new_feature_job_setting)

Done! |████████████████████████████████████████| 100% in 6.1s (0.17%/s)         

Feature Job Setting Analysis Report

Backtest Result

For the feature job setting:
- Period = 86400 s / Offset = 25740 s / Blind spot = 28800 s
The backtest found that all records would have been processed on time.

Done! |████████████████████████████████████████| 100% in 6.1s (0.17%/s)         

Feature Job Setting Analysis Report

Backtest Result

For the feature job setting:
- Period = 3600 s / Offset = 120 s / Blind spot = 240 s
The backtest found that all records would have been processed on time.

In [99]:

# Update Default Feature Job Setting
bureau_table.update_default_feature_job_setting(new_feature_job_setting)

# Update Default Feature Job Setting bureau_table.update_default_feature_job_setting(new_feature_job_setting)

Note:

• This new Default Feature Job Setting will be used by default by any new feature using the table.
• In the Enterprise platform, an approval process is associated with the change in the Default Feature Job Setting and a request to create new versions of existing features is automatically triggered.

Update Default Cleaning Operations at the table level¶

In [100]:

# Get BUREAU table
bureau_table = catalog.get_table("BUREAU")

# Get BUREAU table bureau_table = catalog.get_table("BUREAU")

In [101]:

# Get Info on columns
columns_info = pd.DataFrame(bureau_table.info(verbose=True)['columns_info'])
display(columns_info)

# Get Info on columns columns_info = pd.DataFrame(bureau_table.info(verbose=True)['columns_info']) display(columns_info)

	name	dtype	entity	semantic	critical_data_info	description
0	ClientID	INT	Client	None	None	ID of the client
1	SK_ID_BUREAU	INT	BureauReportedCredit	event_id	None	Recoded ID of previous Credit Bureau credit re...
2	CREDIT_ACTIVE	VARCHAR	None	None	None	Status of the Credit Bureau (CB) reported credits
3	CREDIT_CURRENCY	VARCHAR	None	None	None	Recoded currency of the Credit Bureau credit
4	bureau_application_time	TIMESTAMP	None	None	None	Timestamp when client applied prior applicatio...
5	CREDIT_DAY_OVERDUE	INT	None	None	None	Number of days past due on CB credit at the ti...
6	credit_end_date	TIMESTAMP	None	None	None	When prior CB credit is scheduled to end
7	credit_end_fact	TIMESTAMP	None	None	None	When prior CB credit ended (only for closed cr...
8	AMT_CREDIT_MAX_OVERDUE	FLOAT	None	None	None	Maximal amount overdue on the Credit Bureau cr...
9	CNT_CREDIT_PROLONG	INT	None	None	None	How many times was the Credit Bureau credit pr...
10	AMT_CREDIT_SUM	FLOAT	None	None	None	Current credit amount for the Credit Bureau cr...
11	AMT_CREDIT_SUM_DEBT	FLOAT	None	None	None	Current debt on Credit Bureau credit
12	AMT_CREDIT_SUM_LIMIT	FLOAT	None	None	None	Current credit limit of credit card reported i...
13	AMT_CREDIT_SUM_OVERDUE	FLOAT	None	None	None	Current amount overdue on Credit Bureau credit
14	CREDIT_TYPE	VARCHAR	None	None	None	Type of Credit Bureau credit (Car, cash,...)
15	credit_update	TIMESTAMP	None	event_timestamp	None	Timestamp when did last information about the ...
16	AMT_ANNUITY	FLOAT	None	None	None	Annuity of the Credit Bureau credit
17	available_at	TIMESTAMP	None	record_creation_timestamp	None	Timestamp the record was added to the database

In [102]:

# Get Current Cleaning Operation for Amount column
for info in columns_info.loc[columns_info.name=="AMT_CREDIT_SUM_DEBT"]["critical_data_info"]:
    print(info)

# Get Current Cleaning Operation for Amount column for info in columns_info.loc[columns_info.name=="AMT_CREDIT_SUM_DEBT"]["critical_data_info"]: print(info)

None

In [103]:

# Update Cleaning Operations by handling disguised missing values for AMT_CREDIT_SUM_DEBT
new_cleaning_operations = [
    fb.DisguisedValueImputation(disguised_values=[-99, -98, -96], imputed_value=None),
]
bureau_table["AMT_CREDIT_SUM_DEBT"].update_critical_data_info(
    cleaning_operations=new_cleaning_operations
)

# Update Cleaning Operations by handling disguised missing values for AMT_CREDIT_SUM_DEBT new_cleaning_operations = [ fb.DisguisedValueImputation(disguised_values=[-99, -98, -96], imputed_value=None), ] bureau_table["AMT_CREDIT_SUM_DEBT"].update_critical_data_info( cleaning_operations=new_cleaning_operations )

Note:

• This new Default Cleaning Operations will be used by default by any new feature using the table column.
• In the Enterprise platform, an approval process is associated with the change in the Default Cleaning Operations and a request to create new versions of existing features using the table column is automatically triggered.

Detect features with discrepancies¶

In [104]:

# Define target table
table_name = "BUREAU"
table = catalog.get_table(table_name)
default_job_setting = table.default_feature_job_setting.model_dump()

# Prepare lists for discrepancies
cleaning_discrepancies = []
job_discrepancies = []

# Iterate through all features in the catalog
for feature_name in catalog.list_features()["name"].to_list():
    feature = catalog.get_feature(feature_name)
    feature_info = feature.info()

    # --- Compare cleaning operations ---
    default_cleaning = {}
    actual_cleaning = {}

    # Collect default cleaning operations from the table definition
    for input_col in feature_info["metadata"]["input_columns"].values():
        for col_op in table.column_cleaning_operations:
            col_op = col_op.model_dump()
            if col_op["column_name"] == input_col["column_name"]:
                default_cleaning[col_op["column_name"]] = col_op["cleaning_operations"]
                break

    # Collect actual cleaning operations used by the feature
    for table_cleaning in feature_info["table_cleaning_operation"]["this"]:
        if table_cleaning["table_name"] == table_name:
            for col_op in table_cleaning["column_cleaning_operations"]:
                actual_cleaning[col_op["column_name"]] = col_op["cleaning_operations"]

    # Record discrepancies in cleaning definitions
    if default_cleaning != actual_cleaning:
        cleaning_discrepancies.append(feature_name)

    # --- Compare feature job settings ---
    for job_info in feature_info["table_feature_job_setting"]["this"]:
        if (
            job_info["table_name"] == table_name
            and job_info["feature_job_setting"] != default_job_setting
        ):
            job_discrepancies.append(feature_name)

# --- Summary ---
print("Features with cleaning discrepancies:", cleaning_discrepancies)
print("Features with job discrepancies:", job_discrepancies)

# Define target table table_name = "BUREAU" table = catalog.get_table(table_name) default_job_setting = table.default_feature_job_setting.model_dump() # Prepare lists for discrepancies cleaning_discrepancies = [] job_discrepancies = [] # Iterate through all features in the catalog for feature_name in catalog.list_features()["name"].to_list(): feature = catalog.get_feature(feature_name) feature_info = feature.info() # --- Compare cleaning operations --- default_cleaning = {} actual_cleaning = {} # Collect default cleaning operations from the table definition for input_col in feature_info["metadata"]["input_columns"].values(): for col_op in table.column_cleaning_operations: col_op = col_op.model_dump() if col_op["column_name"] == input_col["column_name"]: default_cleaning[col_op["column_name"]] = col_op["cleaning_operations"] break # Collect actual cleaning operations used by the feature for table_cleaning in feature_info["table_cleaning_operation"]["this"]: if table_cleaning["table_name"] == table_name: for col_op in table_cleaning["column_cleaning_operations"]: actual_cleaning[col_op["column_name"]] = col_op["cleaning_operations"] # Record discrepancies in cleaning definitions if default_cleaning != actual_cleaning: cleaning_discrepancies.append(feature_name) # --- Compare feature job settings --- for job_info in feature_info["table_feature_job_setting"]["this"]: if ( job_info["table_name"] == table_name and job_info["feature_job_setting"] != default_job_setting ): job_discrepancies.append(feature_name) # --- Summary --- print("Features with cleaning discrepancies:", cleaning_discrepancies) print("Features with job discrepancies:", job_discrepancies)

Features with cleaning discrepancies: ['CLIENT_Avg_of_BureauReportedCredits_Available_Credits_104w', 'CLIENT_Avg_of_BureauReportedCredits_Available_Credits_26w', 'CLIENT_Max_of_Active_Cr_active_BureauReportedCredits_AMT_CREDIT_SUM_DEBT_To_AMT_CREDIT_SUMs_104w', 'CLIENT_Max_of_Consumer_credit_Cr_type_BureauReportedCredits_AMT_CREDIT_SUM_DEBT_To_AMT_CREDIT_SUMs_52w', 'CLIENT_Std_of_BureauReportedCredits_Available_Credits_26w']
Features with job discrepancies: ['CLIENT_Avg_of_BureauReportedCredits_Available_Credits_104w', 'CLIENT_Avg_of_BureauReportedCredits_Available_Credits_26w', 'CLIENT_Avg_of_Consumer_credit_Cr_type_BureauReportedCredits_AMT_CREDIT_SUMs_104w', 'CLIENT_Avg_of_Consumer_credit_Cr_type_BureauReportedCredits_End_to_Update_Gaps_104w', 'CLIENT_Max_of_Active_Cr_active_BureauReportedCredits_AMT_CREDIT_SUM_DEBT_To_AMT_CREDIT_SUMs_104w', 'CLIENT_Max_of_Consumer_credit_Cr_type_BureauReportedCredits_AMT_CREDIT_SUM_DEBT_To_AMT_CREDIT_SUMs_52w', 'CLIENT_Std_of_BureauReportedCredits_Available_Credits_26w']

Change feature job setting and cleaning operations of features with discrepancies¶

In [105]:

for feature_name in list(set(cleaning_discrepancies + job_discrepancies)):
    feature = catalog.get_feature(feature_name)
    feature_info = feature.info()
    feature.update_readiness("DEPRECATED")
    
    new_cleanings = []
    if feature_name in cleaning_discrepancies:
        table_cleaning_updated = False
    else:
        table_cleaning_updated = True
    for cleaning_info in feature_info["table_cleaning_operation"]["this"]:
        if not table_cleaning_updated and cleaning_info["table_name"] == table_name:
            new_cleanings.append(
                fb.TableCleaningOperation(
                    table_name=table_name,
                    column_cleaning_operations=table.column_cleaning_operations,
                )
            )
            table_cleaning_updated = True
        else:
            column_cleaning_operations = [
                fb.ColumnCleaningOperation(op)
                for op in cleaning_info["cleaning_operations"]
            ]
            new_cleanings.append(
                fb.TableCleaningOperation(
                    table_name=cleaning_info["table_name"],
                    column_cleaning_operations=column_cleaning_operations)
            )
    if not table_cleaning_updated:
        new_cleanings.append(
            fb.TableCleaningOperation(
                table_name=table_name,
                column_cleaning_operations=table.column_cleaning_operations,
            )
        )
    
    new_feature_job_settings = []
    for job_info in feature_info["table_feature_job_setting"]["this"]:
        if feature_name in job_discrepancies and job_info["table_name"] == table_name:
            new_feature_job_settings.append(
                fb.TableFeatureJobSetting(
                    table_name=table_name,
                    feature_job_setting=table.default_feature_job_setting,
                )
            )
        else:
            new_feature_job_settings.append(
                fb.TableFeatureJobSetting(
                    table_name=job_info["table_name"],
                    feature_job_setting=fb.FeatureJobSetting(**job_info["feature_job_setting"]),
                )
            )
    new_version = feature.create_new_version(
        table_feature_job_settings=new_feature_job_settings,
        table_cleaning_operations=new_cleanings,
    )
    new_version.update_readiness("PRODUCTION_READY")

for feature_name in list(set(cleaning_discrepancies + job_discrepancies)): feature = catalog.get_feature(feature_name) feature_info = feature.info() feature.update_readiness("DEPRECATED") new_cleanings = [] if feature_name in cleaning_discrepancies: table_cleaning_updated = False else: table_cleaning_updated = True for cleaning_info in feature_info["table_cleaning_operation"]["this"]: if not table_cleaning_updated and cleaning_info["table_name"] == table_name: new_cleanings.append( fb.TableCleaningOperation( table_name=table_name, column_cleaning_operations=table.column_cleaning_operations, ) ) table_cleaning_updated = True else: column_cleaning_operations = [ fb.ColumnCleaningOperation(op) for op in cleaning_info["cleaning_operations"] ] new_cleanings.append( fb.TableCleaningOperation( table_name=cleaning_info["table_name"], column_cleaning_operations=column_cleaning_operations) ) if not table_cleaning_updated: new_cleanings.append( fb.TableCleaningOperation( table_name=table_name, column_cleaning_operations=table.column_cleaning_operations, ) ) new_feature_job_settings = [] for job_info in feature_info["table_feature_job_setting"]["this"]: if feature_name in job_discrepancies and job_info["table_name"] == table_name: new_feature_job_settings.append( fb.TableFeatureJobSetting( table_name=table_name, feature_job_setting=table.default_feature_job_setting, ) ) else: new_feature_job_settings.append( fb.TableFeatureJobSetting( table_name=job_info["table_name"], feature_job_setting=fb.FeatureJobSetting(**job_info["feature_job_setting"]), ) ) new_version = feature.create_new_version( table_feature_job_settings=new_feature_job_settings, table_cleaning_operations=new_cleanings, ) new_version.update_readiness("PRODUCTION_READY")

Create new version of a feature list¶

In [106]:

# Get feature list from catalog
feature_list = catalog.get_feature_list("40 features for Credit Default")

# Get feature list from catalog feature_list = catalog.get_feature_list("40 features for Credit Default")

Loading Feature(s) |████████████████████████████████████████| 40/40 [100%] in 0.
Loading Feature(s) |████████████████████████████████████████| 9/9 [100%] in 0.2s

In [107]:

# Check Fraction of default features and production_ready features
print(f"default_feature_fraction: {feature_list.default_feature_fraction}")
print(f"production_ready_fraction: {feature_list.production_ready_fraction}")

# Check Fraction of default features and production_ready features print(f"default_feature_fraction: {feature_list.default_feature_fraction}") print(f"production_ready_fraction: {feature_list.production_ready_fraction}")

default_feature_fraction: 0.825
production_ready_fraction: 0.825

In [108]:

# Create new version
new_version = feature_list.create_new_version()

# Create new version new_version = feature_list.create_new_version()

Loading Feature(s) |████████████████████████████████████████| 40/40 [100%] in 0.
Loading Feature(s) |████████████████████████████████████████| 9/9 [100%] in 0.4s

In [109]:

# Check Fraction of default features of new version
print(f"default_feature_fraction: {new_version.default_feature_fraction}")
print(f"production_ready_fraction: {new_version.production_ready_fraction}")

# Check Fraction of default features of new version print(f"default_feature_fraction: {new_version.default_feature_fraction}") print(f"production_ready_fraction: {new_version.production_ready_fraction}")

default_feature_fraction: 1.0
production_ready_fraction: 1.0

In [110]:

# Check new version is the default
print(
    f"version_name: {new_version.version} \n",
    f"is the new version the new default? {new_version.is_default}"
)

# Check new version is the default print( f"version_name: {new_version.version} \n", f"is the new version the new default? {new_version.is_default}" )

version_name: V251015_1 
 is the new version the new default? True

In [111]:

# Get Default version from Catalog
feature_list = catalog.get_feature_list("40 features for Credit Default")
print(feature_list.version)

# Get Default version from Catalog feature_list = catalog.get_feature_list("40 features for Credit Default") print(feature_list.version)

Loading Feature(s) |████████████████████████████████████████| 40/40 [100%] in 0.
V251015_1
Loading Feature(s) |████████████████████████████████████████| 9/9 [100%] in 0.4s
V240612_1

In [112]:

# List versions
feature_list.list_versions()

# List versions feature_list.list_versions()

Out[112]:

	id	name	version	online_frac	deployed	created_at	is_default
0	68ef6709b5826853a69c9cb8	40 features for Credit Default	V251015_1	0.0	False	2025-10-15T09:19:05.964000	True
1	68ef63333e2dac375b630de7	40 features for Credit Default	V251015	0.0	False	2025-10-15T09:02:51.314000	False

In [113]:

# Check Production Readiness
feature_list.production_ready_fraction

# Check Production Readiness feature_list.production_ready_fraction

Out[113]:

1.0

Delete / Deprecate features and feature lists¶

Only 'DRAFT' features that are not part of a feature list or 'DRAFT' feature list can be deleted. For others, deprecate them instead.

In [114]:

feature = catalog.get_feature("CLIENT_Avg_of_BureauReportedCredits_Available_Credits_26w")

feature = catalog.get_feature("CLIENT_Avg_of_BureauReportedCredits_Available_Credits_26w")

In [115]:

# Check that guardrails prevent you from deleting a feature that is production_ready
try:
    feature.delete()
except Exception as e:
    print(e)

# Check that guardrails prevent you from deleting a feature that is production_ready try: feature.delete() except Exception as e: print(e)

('Only feature with draft readiness can be deleted.', 'Failed to delete specified object.')

In [116]:

# Deprecate it
feature.update_readiness("DEPRECATED")

# Deprecate it feature.update_readiness("DEPRECATED")

Done! |████████████████████████████████████████| 100% in 6.1s (0.17%/s)         
Done! |████████████████████████████████████████| 100% in 6.1s (0.17%/s)         
Loading Feature(s) |████████████████████████████████████████| 2/2 [100%] in 0.2s

In [117]:

# Delete feature list
try:
    feature_list.delete()
except Exception as e:
    print(e)

# Delete feature list try: feature_list.delete() except Exception as e: print(e)

('Only feature list with DRAFT status can be deleted.', 'Failed to delete specified object.')

In [118]:

feature_list.update_status("DEPRECATED")

feature_list.update_status("DEPRECATED")

Concepts in this tutorial¶

• Feature versioning
• Default feature version
• Feature list versioning
• Default Feature Job Setting
• Feature Job Setting Recommendation
• Default Cleaning Operations

SDK reference for¶

• Feature
• FeatureList
• Feature.readiness
• Feature.update_readiness()
• Feature.info()
• EventTable.list_feature_job_setting_analysis()
• EventTable.create_new_feature_job_setting_analysis()
• EventTable.update_default_feature_job_setting()
• FeatureJobSettingAnalysis.backtest()
• TableColumn.update_critical_data_info()
• Feature.create_new_version()
• Feature.list_versions()
• Feature.delete()
• FeatureList.create_new_version()
• FeatureList.list_versions()
• FeatureList.delete()
• FeatureList.production_ready_fraction
• FeatureList.default_feature_fraction

In [ ]:

In [ ]: