MlDomain Metadata Type for Einstein Intent Sets

Architectural Overview#

The MlDomain metadata type represents an Einstein Intent Set—the natural language understanding (NLU) knowledge domain that powers Einstein Bots and conversational service experiences. Rather than hard-coding customer phrases into individual bot dialogs, enterprise architectures employ Intent Sets as reusable, modular natural language models.

By defining MlDomain components in version control, conversational architects and developers manage conversational intents, sample customer utterances, and entity slot classes with standard Git workflows. When deploying updates to production, Intent Sets ensure conversational models remain consistent across multiple bot channels (such as Web Chat, WhatsApp, and SMS).

Directory and File Conventions#

MlDomain files are stored in the mlDomains/ folder in standard Salesforce DX source format.

Key Metadata Fields#

The MlDomain schema encapsulates domain labels, intents, and training data structures:

Child Type: MlIntent#

Declarative XML and Package Manifest Examples#

The following example defines an Intent Set for customer order inquiries:

<?xml version="1.0" encoding="UTF-8"?>
<MlDomain xmlns="http://soap.sforce.com/2006/04/metadata">
    <label>Order Support Domain</label>
    <description>Intent set handling order tracking, status, and shipping inquiries.</description>
    <mlIntents>
        <developerName>Check_Order_Status</developerName>
        <label>Check Order Status</label>
        <mlIntentUtterances>
            <text>Where is my recent shipment?</text>
        </mlIntentUtterances>
        <mlIntentUtterances>
            <text>Track my delivery package</text>
        </mlIntentUtterances>
        <mlIntentUtterances>
            <text>When will my order arrive?</text>
        </mlIntentUtterances>
    </mlIntents>
</MlDomain>

Package Manifest (package.xml)#

Include MlDomain in your deployment manifest:

<?xml version="1.0" encoding="UTF-8"?>
<Package xmlns="http://soap.sforce.com/2006/04/metadata">
    <types>
        <members>*</members>
        <name>MlDomain</name>
    </types>
    <version>68.0</version>
</Package>

CLI Retrieval and Deployment Commands#

Manage MlDomain components with the Salesforce CLI:

# Retrieve all Intent Sets
sf project retrieve start --metadata MlDomain

# Deploy a specific Intent Set to target org
sf project deploy start --metadata MlDomain:Order_Support_Domain

# Inspect existing ML domains via Tooling API
sf data query --use-tooling-api --query "SELECT Id, DeveloperName, MasterLabel FROM MlDomain"

Architect Best Practices and Gotchas#

• Model Retraining: Deploying an updated MlDomain with new utterances updates the metadata definitions in the target org, but you must trigger model retraining (either programmatically or via Setup) to build the active classification model.
• Intent Overlap: Avoid duplicate or semantically identical utterances across different intents within the same domain, as this lowers model confidence and leads to bot fallback errors.
• Bot Dependency: When deploying Einstein Bots, ensure referenced MlDomain intent sets are included in the deployment package to avoid broken intent links.

Frequently asked questions

What does the MlDomain metadata type represent?

It represents an Einstein Intent Set containing intents, training utterances, and slot entity classes used for natural language processing in Einstein Bots.

Where are MlDomain metadata files located in SFDX?

They are located in force-app/main/default/mlDomains/ with the .mlDomain-meta.xml file suffix.

What prerequisite features are required to deploy MlDomain?

The target organization must have Service Cloud, Chat or Messaging, and Einstein Bots enabled before deploying MlDomain metadata.

Can Intent Sets defined in MlDomain be shared across bots?

Yes, MlDomain creates reusable intent sets that can be associated with multiple Einstein Bots across the organization.