SAS Viya Copilot in Action (Part 1): Intelligent Assistance for Building Model Studio Pipelines

[#1]

In my previous post, SAS Viya Copilot Explained: Building Machine Learning Pipelines in Minutes, Not Hours, I introduced SAS Viya Copilot and its core concepts. In this three-part series, we will take a deeper dive into the key capabilities of SAS Viya Copilot for model pipeline development, with a particular focus on how it assists users within Model Studio.

SAS Viya Copilot is designed to make model pipeline development more intuitive, interactive, and efficient. In this post, we walk through the essential ways you can work with SAS Viya Copilot in Model Studio—from submitting prompts and asking general questions to building and running pipelines. You’ll learn how to add specific nodes, explore node details and parameters, and interpret results through automated analyses and summaries. Together, these capabilities show how SAS Viya Copilot helps streamline everyday modeling tasks and accelerates your end-to-end machine learning workflow.

How do you submit prompts to SAS Viya Copilot?

At the bottom of the SAS Viya Copilot side pane is a text field in which you enter your prompts:

Select any image to see a larger version.
Mobile users: To view the images, select the "Full" version at the bottom of the page.

To submit a prompt, enter your prompt in the text field, and then click  or press Enter. To get the best results, ensure that your prompts conform to the best practices listed below.

During your conversation, SAS Viya Copilot maintains a history of your prompts and its responses. You can use the scroll bar to review your conversation history.

How can you get answers to general questions?

If you have general questions about SAS Viya Machine Learning or the nodes in your pipeline, you can prompt SAS Viya Copilot for an answer. These questions typically do not apply to a specific node in the pipeline. Here are some sample questions:

• How do I start building a pipeline in Model Studio?
• What is an unbalanced sample?
• Which model should I try first for my data?
• What are hyperparameters?

The response includes summarized information related to your question. It might also include one or more links to more detailed documentation for your reference.

The questions must be within the scope of SAS Viya Machine Learning. If you ask questions like:

• How is the weather today in Cary?
• What is the distance between New York and New Delhi?

It will result in unsupported responses, as these topics fall outside the capabilities and intended scope of SAS Viya Copilot.

How can you add a specific node to your pipeline?

As you build your pipeline, if you know the node that you want to add next, SAS Viya Copilot can add it for you. If you know the purpose of the next node but are unsure about the node name, SAS Viya Copilot might be able to add an appropriate node for you.

Let me walk you through an example. Once you have created a project and defined the metadata—such as assigning the target variable, setting appropriate roles and measurement levels for variables—and added at least a basic template pipeline containing the required Data node, you are ready to proceed.

SAS Viya Copilot can add a node after an existing node in your pipeline. Currently, it does not support inserting a node before an existing node. SAS Viya Copilot requires you to either select the parent node or explicitly identify it by name in your prompt. In addition, the node you want to add must be a valid child of the selected/identified parent node. In our case, since the pipeline contains only one node, there is no need to select or identify it—the next node will automatically be added below the Data node.

Prompt: Add a node that performs data exploration.

After you submit your prompt, SAS Viya Copilot adds the node to the pipeline. A response similar to the following appears in the side pane:

A new Data Exploration node "Data Exploration" was added after node "Data" in pipeline "Pipeline 1" with default properties.

The response includes a Run node button that you can click to run the newly added node. Alternatively, you can prompt SAS Viya Copilot to run the node.

If you want to add a node and run it immediately afterward, you can combine both actions into a single prompt. For example, when the Data node is selected, this sample prompt adds a Data Exploration node after the Data node, and then runs it:

Prompt:  Add a Data Exploration node and run it.

After you submit the prompt, a response similar to the following appears in the side pane:

A new Data Exploration node "Data Exploration" was added after the Data node in Pipeline 1 with default properties. Now the run process for the Data Exploration node is starting. The node "Data Exploration" in the pipeline "Pipeline 1" has been started. Check the pipeline for status. Next, we will use SAS Viya Copilot to add a data preprocessing node for missing value imputation in a scenario where the parent node is identified by name. Prompt:  Add an Imputation node to the Data node.

After you submit the prompt, a response similar to the following appears in the side pane:

A new Imputation node "Imputation" was added after node "Data" in pipeline "Pipeline 1" with default properties.

Finally, we will use SAS Viya Copilot to add two modeling nodes under different parent nodes using a single prompt.

Prompt: Add a Logistic Regression node to the Imputation node and add a Gradient Boosting node to the Data node.

After you submit the prompt, a response similar to the following appears in the side pane:

A new Logistic Regression node "Logistic Regression" was added after node "Imputation" in pipeline "Pipeline 1" with default properties. A new Gradient Boosting node "Gradient Boosting" was added after node "Data" in pipeline "Pipeline 1" with default properties.

How can you retrieve information about a node?

If you need more information about a node, you can prompt SAS Viya Copilot to provide a description of the node. When you select a node and open the SAS Viya Copilot side pane, one or more buttons are provided in the side pane that allow you to view information about the selected node or perform specific actions. In the pipeline created so far using SAS Viya Copilot, I got the Gradient Boosting node selected and opened the SAS Viya Copilot. If the Gradient Boosting node has not yet been run, you will see the options shown on the left. Once the node has been run, the option buttons shown on the right become available.

You can also prompt SAS Viya Copilot for information about any node in your pipeline. In order to prompt for information about a node, you must select the node before you submit your prompt, or you must identify it by name in your prompt. Here are some sample prompts:

• For a node that is selected:
  • What is this node?
  • How is this node used?
• For a node that is identified by name:
  • What is the Logistic Regression node?
  • How is the Imputation node used?

SAS Viya Copilot provides a description of the selected node in the side pane. The description might also include links to more detailed documentation for your reference.

How can you view details about node parameters?

As you adjust the parameters of the nodes in your pipeline, if you need information about a particular parameter and its possible values, you can prompt SAS Viya Copilot for information.

I am continuing with the pipeline created so far using SAS Viya Copilot, with the Gradient Boosting node selected. Below are some sample prompts I can use to get information about gradient boosting hyperparameters:

• What does the Learning Rate property do?
• What values can I use for L1 and L2?
• What is LightGBM?

SAS Viya Copilot displays information about the parameter in the side pane. The information might include one or more links to more detailed documentation for your reference.

How can you run a node?

When SAS Viya Copilot adds one or more nodes to your pipeline, it displays a button in the side pane that lets you run the newly added node(s) directly. In other situations, you can also ask SAS Viya Copilot to run a node for you. To do this, either select the node before submitting your prompt or explicitly refer to the node by name in your prompt. Below are some sample prompts based on the existing pipeline context:

• For a node that is selected:
  • Run node.
  • Run this node. 
• For a node that is identified by name:
  • Run the Logistic Regression node.
  • Run Gradient Boosting.

Here is an example of the response in the SAS Viya Copilot side pane:

The node "Logistic Regression" in the pipeline "Pipeline 1" has been started. Check the pipeline for status.

SAS Viya Copilot indicates in its response the name of the node that is running and the name of the pipeline that contains the node. Any preceding nodes that have not been run are run before the selected or specified node is run. SAS Viya Copilot does not indicate whether the node ran successfully. You must monitor the status of the node or nodes in the pipeline as they run to determine whether they run successfully.

How can you get an analysis or summary of node results?

After you run a node, you can prompt SAS Viya Copilot to provide an analysis or summary of the node results. To get the analysis, you must first select the node or you must identify the node by name in your prompt. Here are some sample prompts for getting an analysis of a selected node based on the existing pipeline context:

• For a node that is selected:
  • Explain the results.
  • What do these results mean in practical terms?
• For a node that is identified by name:

  Describe the Gradient Boosting node results.

SAS Viya Copilot provides an analysis or summary of the results in the side pane. It includes a high-level overview of the modeling node or data preprocessing node, an analysis of the key findings, and a conclusion. Additional information is provided based on the model type and node parameter settings.

For example, the explanation covers how the Gradient Boosting model works, what it predicts, and how results are interpreted using probability scores and cutoffs. It explains model behavior through variable importance, ROC/KS analysis, and detailed fit statistics across training and validation partitions. Together, these elements provide insight into key drivers of predictions, overall model accuracy, discrimination power, generalization, and areas for potential improvement.

Stay tuned and watch for the upcoming posts in this series:

• SAS Viya Copilot in Action (Part 2): Smart Recommendations for the Next Nodes in Your Model Studio Pipeline
• SAS Viya Copilot in Action (Part 3): Improving and Optimizing Nodes in Your Model Studio Pipeline

Find more articles from SAS Global Enablement and Learning here.