​

Intro

AI Models and AI Projects are the places to store and manage models in Domo. You can store models of different types—like large language models (LLMs) and data science and machine learning (DSML) models—and from various internal and external sources, including Domo Jupyter Workspaces, OpenAI, and Hugging Face. AI Models lets you review the details of your AI models in Domo and compare their performance. From this storage space, you can deploy your AI models to different parts of Domo, including Magic ETL, where the AI Model Inference tile uses your models to transform data. Learn more about the AI Model Inference tile. AI Projects allows you to create new projects that contain existing models. A project is a collection of models that allows you to compare and contrast different models and quickly visualize different model metrics. Tip: Before adding a model to a project, the model must be added to your Domo instance. Learn how to add a model below.

---

• Required grants
• Access AI Models and AI Projects

AI Models

• Add a model
  • Add a third-party model
• View model information
  • Edit the schema
  • View schema properties
  • Edit schema
  • Add new input/output
  • Delete input/output
• Work with model endpoints
  • Start and stop endpoint
  • Test endpoint
  • View endpoints logs
• Share access to a model
• Deploy a model to Magic ETL
• Delete a model

AI Projects

• View project information
• Create a project
  • Create a project using AutoML
  • Create a custom project
• Add a model to AI Projects
• Compare models

---

​

Required Grants

To access AI Models and AI Projects, both of the following grants must be enabled for your role:

• Create AI Service Models — Allows you to create new models and edit, deploy, and delete models you can access.
• (Admin-only) Manage AI Services — Allows you to create, edit, deploy, and delete all models and projects in AI Models.

Learn about adding grants to custom roles.

​

Access AI Models and AI Projects

From the Domo navigation, select More > Domo AI to open the Domo AI interface. Use the left navigation to go AI Models or AI Projects.

​

Add a Model

You can add models to AI Models using Domo’s internal tools or third-party sources.

​

Add a Third-Party Model

Follow these steps to add a model from a third-party source:

1. Access AI Models.
2. On the AI Models page, select Add Model.
   
   The model import modal displays.
   
3. Complete the following steps outlined in the modal. After completing a step, select Next Step to move forward. At any time, you can select Back to return to the previous step or Cancel to exit the modal.
   1. Select an Adapter — The adapter is the third-party source of the model.
   2. Select an Account — You or your company must have an account with the third-party source of the model that is integrated with Domo.
      • If you haven’t already integrated an existing third-party source with Domo, select Create Account and complete account integration. Learn how to integrate an account. After integrating your account, restart the Add a Model process.
   3. Import a Model — Choose the model to import from the list of available models from that source.
   4. Configure Tasks — Specify the following:
      • Select the model task from the Task dropdown, which lists general model capabilities. The model account determines the available tasks. Possible task types include Binary Classification, Chat Completion, Classification, Image Generation, Regression, Summarization, Task Planning, Text Embedding, Text Generation, and Others.
      • Select input or output from the Schema dropdown.
      • Choose either CSV or JSON from the Media Type dropdown. In the image below, the task type is Chat Completion, the schema is input, and the media type is JSON
        
        This section of the modal includes a preview of the model metrics and hyperparameters, which are brought in from the selected adapter account.
      • (Optional) Select + Add New to add a new input/output value row to your schema.
        
        View in Code Editor: Toggling the View in Code Editor switch displays the backend code that defines the model.

        

        You can edit the code directly from the Code Editor to automatically update the model schema. Select Update Schema to save your changes.

        
   5. Configuration — Fill out the required information. The following notes may be helpful:
      • The model task is auto-selected based on the adapter you selected in step 1 of this modal. You can override the selected option by selecting a new task type from the dropdown.
      • The Max Tokens are the maximum number of tokens the model can process in a single interaction (input prompt + output). A token is a sequence of characters used to identify the model.
      • (Optional) Selecting the checkbox labeled Add to a project after creation opens the Manage Projects modal, where you can choose projects to which to add the model. Learn more about AI Projects below.

Select Import Model to complete the setup. Your new AI model now displays on the AI Models page, and y ou can securely use it to complete tasks in Domo.

​

Review Model Information

On the AI Models page, you can view the following about each model you have access to:

• The model name, description, owner, and the date it was created.
• The number of projects the model is added to.
• The current status of the model endpoint. See Work with Model Endpoints to learn more.
• Metrics associated with the model. You can see more about the metrics in the Details view.
  Note: When filtering the models, the Host Context Type filter refers to the source of a model.

  

Selecting a model from the AI Models page opens the Details view for that model. The Details view includes four sections, described in the table below. The values in each section depend on the model source and how it was trained and brought into Domo.

Section	Description	Image
Metrics	The values displayed here for metric name, value, and standard deviation are created during the initial training of the model when it is brought in from a third-party source. The model values and other results are initially compared to the sample data used in the model training. The Standard Deviation column displays even when your model doesn’t have them as an output.
Hyperparameters	These values are created during the initial model training and typically include items like limits, top values, variants, and any model configuration done external to Domo.
Model Information	These values include the machine learning algorithm the model was trained on, the model ID used to run the model endpoint in other areas in Domo, snapshot data if a snapshot exists, and a name and link to the model job (if the model was created in Domo).
Model Schema	These values include the model’s assigned task, the input type, input schema, output type, and output schema. Select Edit Schema for both the input and output to update them in the schema. Learn how to edit the schema.

​

Edit the Schema

From the model Details view under Model Schema, you can select Edit Schema for both the input and output schema to make changes to them in the Edit Schema modal. In the modal you can view properties of the schema and make the following edits: edit rows, delete rows, or add a new row. Select Update Schema to save your changes.

​

View Schema Properties

For any row in the schema, select Action menu > View Properties to see the description, values, and array properties for each input. These properties are based on your model schema and how the model was trained.

​

Edit Schema

For any row in the schema, select Action menu > Edit to change the name, default value, data type, description, and values for that input/output. The available data types are Array, Boolean, Date, Datetime, Decimal, Double Long, Numeric, Object, and String Types. The accepted input and output values rely on your model schema and how the data was trained.

​

Add New Input/Output

In the Edit Schema modal, selecting + Add New allows you to add a new input/output value row to your schema and edit its properties.

​

Delete Input/Output

For any row in the schema, select Action menu > Delete to remove the input/output from the list. If you then select Update Schema, the input/output is permanently deleted. This action cannot be undone.

​

Work with Model Endpoints

Endpoint information and the testing interface are available in the Endpoint tab of the model details.

​

Start and Stop Endpoint

Starting and stopping the endpoint from the Endpoint tab also starts and stops it in other places it’s being used in Domo, like Magic ETL. The endpoint status displays next to the St art/Stop Endpoint control.

​

Test Endpoint

The Test Endpoint section of the Endpoint tab allows you to enter input and output text and run a test on your model. You can also copy the test output, view the schema, and view the model code snippet for the endpoint test. Follow these steps to edit the code snippet used for the model:

1. Select Code Snippet to open a modal that displays the model code snippet.
2. Select the code language from the dropdown.
3. Select Copy Code to copy the code snippet to your machine.
4. To update the code, make changes by directly modifying the code box.
   

​

View Endpoint Logs

The Endpoint Logs tab is only visible if you have started or are currently running model endpoints. In the tab, you can see event info for any model endpoints that have run or are running. Possible endpoint states include: Started, Running, Stopped, Failed, Active, and Snapshot Failed. The current endpoint status displays on the AI Models page.

​

Share Access to a Model

From the AI Models page, select Action menu > Manage Permissions. Use the dropdown to assign people/groups one of the following levels of access:

• Can Execute — Grants access to view and execute the model.
• Can Share — Grants access to view, execute, and share the model with others in your instance.
• Admin — Grants access to view, execute, share, delete, and update the model.
  

​

Deploy a Model to Magic ETL

Deploy a model to Magic ETL from the AI Models page by selecting Action menu > Deploy to ETL. The Magic ETL canvas opens with an AI Model Inference tile deployed. Learn more about the AI Model Inference tile.

​

Delete a Model

You can delete models you have access to from the AI Models page by selecting Action menu > Delete. This action cannot be undone.

​

View Project Information

On the AI Projects page, you can view the name, description, owner, creation date, and date of last update for every project you can access. Select a project from the list to open its Details view, where you can compare models. If you don’t yet have access to any project, learn how to create a new project below.

​

Create a Project

You can add projects to AI Projects using the Domo AutoML feature or create a custom project. Learn more about Domo AutoML.

​

Create a Project using AutoML

Follow the steps below to create a new project using Domo AutoML:

1. Access AI Projects.
2. Select New Project to open the Create new project modal.
3. In the modal, select Redirect to AutoML.
   
4. Use the dropdown to select a DataSet to create your project on and select Go To DataSet. The AutoML tab of the DataSet opens.
   Important: To run an AutoMl job, your project DataSet must have at least 500 rows of data.
   
5. Select Get Started.
   
6. Select the column from your DataSet you want to predict and the task type. There are four types of tasks:
   • Automatic (Recommended) — This type uses the best task based on your data.
   • Regression — This type measures numeric values such as revenue.
   • Binary Classification — This type classifies boolean data such as customer churn.
   • Multi-class Classification — This type classifies data groups such as industry, tech, and finance.
   
7. Select Start Training. A timer displays while AutoML trains and tests learning models using your data. Learn more about the stages of training.
   

When AutoML is finished, a Model Overview page displays information about the performance of the models that AutoML built. Your AutoML project now displays on the AI Projects page. Learn more about preparing data for AutoML and training and deploying models with AutoML.

​

Create a Custom Project

Follow the steps below to create a new custom project:

1. Access AI Projects.
2. Select New Project to open the Create new project modal.
3. In the modal, select Custom Project.
   
4. Name your project and enter an optional description.
5. Select + Add Existing Models to add a model to the project.
   
6. Select Create New Project to add the project to the AI Projects page.

Learn more about managing AI models.

​

Add a Model to AI Projects

From the AI Models page, you can add a model to one or more projects by selecting Action menu > Manage Projects to open the Manage Projects modal. In the modal, you can add or remove the model from existing projects and save your changes.

​

Compare Models

When you select a project from the AI Projects list, you open the project’s Details view. Here, you can compare up to 10 models and 10 metrics total. The available metrics depend on the models brought into your project and how they were initially trained. A project’s Details view contains the Metric Comparison section, where you can view the metrics you are comparing, and the model list, where you can see b asic information about each model in the project. Select a model to see its additional details.

​

Choose Models to Compare

In the model list, check the box for each model you want to compare (you can choose up to 10). Then, choose metrics to compare under Metric Comparison. When you have at least two metrics selected, you can compare multiple metrics within a single model, the metric comparison loads as side-by-side comparison bar graphs. In the example above, the AutoML-generated model is used, and the validation:accuracy, validation:macro\_recall, and validation:mlogloss metrics are compared side-by-side.