Deployment

See also

UI Tutorial: Deploy and Serve | Concepts: Deployment | API Tutorial: Credit Default — Step 13 | API Tutorial: Store Sales Forecast — Step 8

Prerequisites

This page uses the wait_for_task helper defined in API Overview.

Once you have a trained model and a feature list, you can deploy them to serve predictions in production.

Create a Deployment

import featurebyte as fb

client = fb.Configurations().get_client()

response = client.post(
    "/deployment",
    json={
        "name": "Daily Sales Forecast - Production",
        "feature_list_id": feature_list_id,
        "ml_model_id": ml_model_id,
        "use_case_id": use_case_id,
    },
)
task_id = response.json()["id"]
task = wait_for_task(client, task_id)

deployment_id = task.get("payload", {}).get("output_document_id")

Parameters:

Parameter	Type	Required	Description
name	string	Yes	Display name for the deployment
feature_list_id	string	Yes	ID of the feature list to deploy
ml_model_id	string	Yes	ID of the trained model to deploy
use_case_id	string	Yes	ID of the use case this deployment serves

Enable or Disable a Deployment

A newly created deployment is disabled by default. Enable it to start serving predictions:

# Enable
response = client.patch(
    f"/deployment/{deployment_id}",
    json={"enabled": True},
)
task_id = response.json()["id"]
wait_for_task(client, task_id)

# Disable
response = client.patch(
    f"/deployment/{deployment_id}",
    json={"enabled": False},
)
task_id = response.json()["id"]
wait_for_task(client, task_id)

List Deployments

response = client.get(
    "/catalog/deployment",
    params={"page": 1, "page_size": 20},
)
deployments = response.json()["data"]

Response fields (each item in data):

Field	Type	Description
id	string	Deployment ID
name	string	Deployment name
feature_list_id	string	Deployed feature list ID
ml_model_id	string	Deployed model ID
use_case_id	string	Associated use case ID
enabled	boolean	Whether the deployment is currently active

Generate Deployment SQL

Generate SQL queries that can be scheduled in your data warehouse to compute features:

response = client.post(
    "/deployment_sql",
    json={"deployment_id": deployment_id},
)
task_id = response.json()["id"]
wait_for_task(client, task_id)

response = client.get(
    "/deployment_sql",
    params={"deployment_id": deployment_id},
)
sql = response.json()

Parameters:

Parameter	Type	Required	Description
deployment_id	string	Yes	ID of the deployment to generate SQL for

Response fields (GET /deployment_sql):

Field	Type	Description
deployment_id	string	ID of the deployment
sql_queries	array	List of SQL query strings to execute in your warehouse

The generated SQL can be used with your warehouse's scheduling system (e.g., Snowflake Tasks, dbt, Airflow) to compute features on a schedule without depending on the FeatureByte server at inference time.