Archived models

When inspecting the Status column of the Models (platform-ca | platform-eu | platform-au) page of the Coveo Administration Console, you may notice that some of your Coveo Machine Learning (Coveo ML) models have the Archived status, or contain a warning message stating that the model will be archived soon.

A Coveo ML model is automatically archived when it hasn’t been queried for at least 30 days. In other words, if a model hasn’t been requested through a Search API call in 30 days the model falls into a state in which it can no longer instantly answer user queries.

  • The Status column of a given model displays a warning 10 days before being archived. This warning helps you take action before the model is archived.

    warning for archived models as shown in the Coveo Administration Console
  • Archived models count against the number of models granted in your Platform plan.

An archived model no longer affects search results or improves the overall user experience. Therefore, you should consider deleting it so that the models that appear on the Models (platform-ca | platform-eu | platform-au) page are only those used on your production interfaces.

However, if you want to reactivate the model, see Reactivate Archived Models for instructions.

Delete archived models


You must dissociate the model from all its associated query pipelines before deleting it.

  1. On the Models (platform-ca | platform-eu | platform-au) page, click the ML model that you want to delete, and then click More > Delete in the Action bar.

  2. In the Delete a Model panel that appears, click Delete model.

Reactivate archived models

You can reactivate archived models if needed.

For an archived model to become active again, the model must be queried at least once.


When an archived model is queried, the model initiates a rebuild if necessary. This ensures that the model only learns on the most recent data and doesn’t provide outputs based on the data learned before the model was archived.

If a rebuild occurs, the model falls into the Building status, and won’t provide any output until the build process is complete.

The model will start providing outputs when it falls into the Active status.

To be queried, the model must first be associated with a query pipeline to which queries are directed. Then, the Search API must evaluate and execute this model-pipeline association.


For Case Classification (CC) models, which can’t be associated with a query pipeline, models must be associated with a Case Assist configuration and requested by the Customer Service API to be queried.

The Search API evaluates models listed in the Machine Learning section of a query pipeline configuration in the order in which they appear on the list.

See Order of execution of query pipeline features and Change the rule order for more information about how model-pipeline associations are evaluated and executed by the Search API, and on how to change the order of a given model-pipeline association.