Skip to main content

Create and Manage Workload Profiles

Workload profiles are reusable templates that define how an application should be configured, decoupling the modeling and deployment phases of the application lifecycle. You build workload profiles from definitions (components, traits, and policies) and use them when creating profile bundles and deploying applications and models with those profile bundles.

Create Workload Profiles​

Create a new workload profile to define a reusable application, model, or infrastructure template.

Prerequisites​

Enablement​

  1. Log in to PaletteAI.

  2. From the left main menu, select Workload Profiles.

  3. In the top-right, select New Profile. Alternately, beside an existing workload profile, select the three-dot menu, and choose Clone.

  4. Enter the Basic Information for your workload profile. The available fields depend on whether your PaletteAI instance uses basic or semantic versioning. The following tables describe the fields for each mode. Select Next when finished.

    info

    Versioning mode is a system-wide setting configured during installation via the Helm value global.featureFlags.versioningType. The default is basic (v0, v1, ...). Set to semantic for semantic versioning (v0.0.1, v0.1.0, ...). Refer to Helm Chart Configuration for details.

    Basic Information​

    ParameterDescriptionRequired
    NameA unique name for your workload profile within the project scope. Must start and end with alphanumeric characters, and can only contain lowercase letters, numbers, hyphens, and periods.✅
    DescriptionA description for your workload profile.❌
    Profile TypeThe type of workload profile: Application, Model, or Infrastructure. Determines which profile bundles can include this workload profile. Refer to Workload Profile Types for details.✅
    New RevisionThe revision identifier. Automatically incremented and cannot be modified manually.✅
    Revision notesNotes describing what changed in this revision.❌

  5. Use the Create Profile screen to build your workload profile stack using components, Helm charts, and traits. The profile stack is a visual representation of the layers that make up your AI/ML application or model.

    Refer to the applicable section based on what you would like to add to your workload profile.

    Drawer Actions​

    The following table describes the actions available for individual parts of the workload profile stack and which layers support each action. The numbers in the table correspond to the numbers in the image.

    Component drawer actions

    NumberOptionDescriptionComponentsTraits
    1PriorityChange the component's deployment priority.✅❌
    2Form and YAML editorUse the hamburger icon to complete information using the form editor. Use the </> icon to switch to the YAML editor.✅✅
    3Trash iconRemove the selected component, Helm chart, or trait from the stack.✅✅
    4{} iconToggle macro mode to enter values using {{variable}} syntax. Macros can reference definition outputs, object outputs, system outputs, and user variables.✅✅
    5Arrow iconRevert a field to its default value. Only available for certain fields.✅✅
    6Show Optional PropertiesReveal additional, optional configurable fields.✅✅
    7Errors and Definition outputsView validation errors and add definition outputs as necessary.✅✅

    Add Components​

    Components are the main building blocks of your workload profile. Each component represents a discrete unit of your application, such as a web server or a worker process.

    1. Select Add Component or a layer in the stack to open the Add Component dialog.

    2. Browse for the component you want to add. Choose a component, version, and select Add Component.

      tip

      Use the Show system definitions toggle to control which components are displayed. When enabled (the default), both system-provided and user-created components are shown. When disabled, only user-created components within your project scope are displayed, filtering out components in the mural-system namespace.

    3. The component is added to the stack and a Component drawer opens for configuration. The drawer displays only required fields by default. Configure the component as necessary using the actions in the Drawer Actions table.

    4. Close the drawer once all required fields have been configured.

    Add Helm Charts​

    You can add Helm charts to your workload profile from Helm repositories or OCI repositories. When a Helm chart is added to a profile, it is automatically converted to a component.

    1. Select Helm Chart to open the Helm chart dialog. 6. Choose between the Helm Repositories and OCI Repositories tabs.
      • Helm Repositories — Select a repository from the list, then choose a chart and version from the available options.
      • OCI Repositories — Select a repository from the list. The chart name and version are automatically extracted from the repository's URL and reference (tag, semver, or digest).
    2. Select Add to add the chart. The chart's top-level items are added to the stack as components, and the Component drawer opens for configuration. Required and optional fields are automatically sorted, and default values are pulled in where applicable. Configure the component as necessary using the actions in the Drawer Actions table.
    3. Close the drawer once all required fields have been configured.

    Add Traits​

    Traits modify how a component runs, such as scaling behavior or health checks. Traits are added per-component and are only visible when their parent component is selected in the stack.

    1. Select a component in the stack.

    2. Browse for the trait you want to add. Choose a trait, version, and select Add Trait.

      tip

      The trait selection modal also includes a Show system definitions toggle. When disabled, only user-created traits are shown.

    3. The trait is added to the component and a Trait drawer opens for configuration. Configure the trait's required fields. Configure the trait as necessary using the actions in the Drawer Actions table.

    4. Close the drawer once all required fields have been configured.

    Add Object Outputs​

    Object outputs expose metadata from Kubernetes resources that exist outside the workload. 1

    1. Select Object Outputs to open the Object Outputs drawer.

    2. Select Add Object Output to open the form. Configure the following fields as necessary.

      ParameterDescriptionRequired
      Output nameA unique identifier for this output.✅
      ScopeWhether the referenced resource is Cluster-scoped or Namespace-scoped.✅
      NamespaceThe namespace of the referenced resource. Only displayed when Scope is set to Namespace.✅
      API VersionThe apiVersion of the referenced Kubernetes resource.✅
      KindThe kind of the referenced Kubernetes resource.✅
      Resource NameThe name of the referenced Kubernetes resource.✅
      PathA JSON path to a field under metadata, spec, or status of the referenced resource.✅
      Skip HashWhether to exclude this output from the workload hash. Defaults to false.❌
      Deployment PriorityAssign the output to a specific deployment priority, or leave unset. Unscoped outputs are resolved only after the final priority is processed.❌

    3. Select Add to save the object output.

    4. Use the Add Object Output button to add additional outputs as needed. Close the drawer when finished.

  6. Add as many components, Helm charts, traits, and object outputs as necessary. Once you have finished building your profile stack, select Next.

  7. On the Review screen, verify and finalize your workload profile configuration. If changes are needed, use the step navigation on the left to return to the applicable screen. When satisfied, select Save Profile.

Validate​

  1. Log in to PaletteAI.

  2. From the left main menu, select Workload Profiles.

  3. Verify your new workload profile is listed in the table. The columns display the profile name, type, description, and latest revision or version.

Edit Workload Profiles​

Edit an existing workload profile by creating a new version or revision. Each edit preserves the previous version, enabling you to track changes over time.

If a workload profile is part of a profile bundle, you can edit your profile bundle use the updated version of your workload profile.

Prerequisites​

Enablement​

  1. Log in to PaletteAI.

  2. From the left main menu, select Workload Profiles.

  3. Locate the workload profile you want to edit. Select the three-dot menu and choose View Details.

  4. On the profile detail page, you can review the profile's configuration in a read-only view. Use the version drop-down in the breadcrumb to switch between existing versions.

  5. Select Create Version (semantic versioning) or Create Revision (basic versioning) to open the edit wizard.

  6. Update the Basic Information as needed. When editing, the Name and Profile Type fields are read-only. You can modify the Description, version notes, and (if semantic versioning is enabled) the Update type. Refer to the Basic Information table for additional details. Select Next when finished.

  7. On the Create Profile screen, modify the workload profile stack as needed. You can:

    • Add, remove, or reconfigure components and traits
    • Add or remove Helm charts
    • Add, edit, or remove object outputs and definition outputs
    • Modify field values, deployment priorities, and macro references

    Refer to the drawer actions table and applicable component, Helm chart, trait, and object output section for additional details.

    Select Next when finished.

  8. On the Review screen, verify your changes. Update Annotations and Labels as needed.

  9. Select Save Profile to save the new version.

Validate​

  1. Log in to PaletteAI.

  2. From the left main menu, select Workload Profiles.

  3. Verify the workload profile's Latest revision or Latest version column reflects the new version number.

Delete Workload Profiles​

Remove a workload profile that is no longer needed. Deletion is permanent and cannot be undone.

Prerequisites​

Enablement​

  1. Log in to PaletteAI.

  2. From the left main menu, select Workload Profiles.

  3. Locate the workload profile you want to delete. Select the three-dot menu and choose Delete. The workload profile is deleted immediately.

Validate​

  1. Log in to PaletteAI.

  2. From the left main menu, select Workload Profiles.

  3. Verify the workload profile is no longer listed in the table.

Next Steps​

Once you have a workload profile, you can use it to create a profile bundle to deploy an AI/ML application or model.