Managing Query Pipelines
- What Is a Query Pipeline
- Routing Mechanisms and Rules
- Deployment Overview
- Adding Coveo ML Automatic Relevance Tuning Models
- Adding Coveo ML Query Suggestions Models
- Adding Coveo ML Recommendations Models
- Editing a Coveo ML Model Without Service Interruption
- Managing Thesaurus
- Managing Featured Results
- Managing Stop Words
- Managing Ranking Expressions
- Managing Ranking Weights
- Managing Triggers
- Managing Filters
- Managing Query Parameters
- Managing Rules and Rule Conditions
- Coding Query Pipeline Rules
- Leveraging Custom Context Information in Salesforce
- Testing Changes
- Exporting/Importing Between Organizations
Editing a Coveo Machine Learning Model Without Service Interruption
You can edit an existing Coveo™ Machine Learning (Coveo ML) model (for Automatic Relevance Tuning [ART] or Query Suggestions) that is in Ready state, but currently however, the service supplied by a Machine Learning model will be interrupted for about one hour while the model is trained with the new configuration (see Coveo Machine Learning Models).
You want to change the configuration of your Query Suggestions model. After changing at least one parameter, once you click Apply in the Edit a Machine Learning Model panel, no suggestions will appear below the search box using this model in this pipeline for about one hour.
While a service interruption may be acceptable in a developer or staging environment, it is obviously not in a production environment.
The work around is to rather create a new model with the desired configuration, wait for the new model to be ready, and then make the new model active, as described in detail in the following procedure (see Who Can Perform the Page Actions).
Duplicate and Swap Coveo Machine Learning Models Without Service Interruption
If not already done, log in to your Coveo Organization with an Admin or Relevance Analyst user role account.
In the navigation bar on the left, under Search Optimization, select Query Pipelines.
In the Query Pipelines page, click the Name of the pipeline containing the Coveo ML model that you want to duplicate and modify.
In the selected pipeline page, select the Machine Learning tab.
Duplicate and modify the duplicated model:
In the Machine Learning tab, at the right end of the line for the model you want to duplicate, click the Actions icon (), and then select Duplicate.
A line for the new model appears at the bottom of the model list and in the Status column, In progress appears for a few seconds before showing Ready.
Because the first model in the list matching your query is applied, your original model is still providing the service, not the duplicated model.
At the right end of the line for the duplicated model, click the Actions icon (), and then select Edit model.
In the Edit a Machine Learning Model panel, make the desired changes to your Coveo ML model configuration, and then click Save (see Adding Machine Learning Automatic Relevance Tuning Models, Adding Machine Learning Query Suggestions Models, or Adding Coveo Machine Learning Recommendations Models).
Back in the selected pipeline page on the new model line, within a few seconds the Status column shows In progress while the model is being trained for the new configuration.
Wait for new model Status column to show Ready, typically about one hour.
Swap the models: in the list of models, drag-and-drop the new model line above the original model line.
Because the new model now appears before the original one, it starts providing the service for incoming matching queries, no longer the original one.
Validate that the service is effectively supplied by the new model.
When you plan to never use the original model again, it is a good practice to delete it by clicking the delete icon () at the end of the model line.
Who Can Perform the Page Actions
The required roles are: