Skip to content

11. Deploy and Serve a Feature List

You'll designate features as Production-Ready, navigate FeatureByte's Approval Flow, and once all features in your list receive approval, deploy it for both online and batch serving.

Step 1: Mark a Feature as Production-Ready

  1. Navigate to the Feature catalog from the Experiment section of the menu. Locate the most predictive feature by sorting based on the predictive score, here "CUSTOMER_Count_of_invoices_13w". Name

  2. Click the three dots next to the feature's current status (e.g., DRAFT) Readiness Tree Dots and change its readiness to Upgrade Icon. Name

Note

If your catalog has an Approval Flow enabled, upgrading features to "Production-Ready" requires approval.

Step 2: Enable Approval Flow

  1. Click on the settings icon Settings icon located next to the Catalog name at the top of the screen. Name

  2. Inside the catalog settings, locate and enable the 'Approval Flow' option. Save changes. Name

  3. Look for a validation mark next to the Catalog name. This indicates the Approval Flow is enabled. Name

Step 3: Request Production-Ready Upgrade

  1. Navigate to the Feature catalog and search "CUSTOMER_Count_of_items_by_product_ProductGroup_13w" feature in the Feature Catlog, click Upgrade Icon and confirm you want to change its readiness. Name

  2. This initiates an approval request. Click View Button to see the request. Name

  3. This will lead you to 'New Production-Ready Feature' under 'Feature Requests' of the 'Approve' section of the menu. Name

Step 4: Review Checks

Click on the request to view details and checks.

Note

Checks include several evaluations, such as checking the feature's compliance with default cleaning operations and the feature job setting of its source tables. It also involves confirming the status of these tables and backtesting the feature job setting to prevent future training-serving inconsistencies.

Name

If any checks fail, resolve them. In our case, a check is failing because no backtest has been run to assess the feature job setting.

  1. Click Run Analysis Button to run a "Data Availability and Freshness Analysis" if the check failed.
  2. Once this is done, click Run Backtest Button to run a backtest of the feature job setting.

This should turn the checks green.

Name

Step 5: Review Feature

Review the feature details, including its definition and use of tiles (partial aggregations).

Name

Name

Step 6: Request Review

Once the request is ready for review, request it by clicking Request Review Button.

The request is now awaiting approval.

Name

Step 7: Approve and Merge Request

After review, approve it by clicking Approve Request Button.

Name

Merge it by clicking Merge Button.

Step 8: Audit Merged Request

Click the 'Merged' tab to see details of merged requests.

Name

Step 9: Requests by batch

To mark all features in a feature list as "Production-Ready",

  1. Go to the Feature List Catalog.
  2. Click Check Button next to Calculator Button of the feature list you want to upgrade.

Name

Name

Confirm.

Name

This will trigger a pop-up window. Click View Button.

Name

This will lead you to 'New Production-Ready Feature' under 'Feature Requests' of the 'Approve' section of the menu.

Since all checks passed, select all requests by checking the box next to 'Name', and then request reviews in batches.

Name

Approve requests.

Name

Merge requests.

Name

The merged requests can be found in the Merged tab.

Name

Step 10: Deploy

Now that all features in our feature list are "Production-Ready," we can deploy it.

  1. Click Rock Icon to initiate deployment. Name

  2. Provide a name (e.g. "Simple set for Customer Activity next Week @ store") and associate the deployment with your Use Case for tracking. Name

  3. Click 'View Deployments' to access the deployment in the Deployment Catalog. Name

Step 11: Enable Deployment

In the Deployment Catalog (under the 'Manage' Section), locate the new deployment and select it.

Name

Name

Click the three dots next to the 'NEW' label and choose 'Enable.'

Name

Name

Step 12: Batch Serving

You can perform batch serving using the SDK or by navigating to the "Batch Feature Tables" tab. There, you can generate a new Batch Feature Table by selecting a source table that contains the entity IDs relevant to your use case.

Name

Although batch serving is available in this tutorial, online serving appears to be the most appropriate approach for the use case.

Step 13: Online Serving

In the 'Online Serving' tab, find Python or shell script templates for REST API services.

Name

Name

Go to the 'Preview' tab to see examples of feature values that will be served.

Name

Use the SDK for batch serving if needed.