> ## Documentation Index
> Fetch the complete documentation index at: https://docs.open-metadata.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Create an ML Model Service

> Create a new ML model service connection

# Create an ML Model Service

Create a new ML model service connection to a platform such as Mlflow, Sklearn, or SageMaker.

## Body Parameters

<ParamField body="name" type="string" required>
  Name of the ML model service. Must be unique across all ML model services.
</ParamField>

<ParamField body="serviceType" type="string" required>
  Type of ML model service (e.g., `Mlflow`, `Sklearn`, `SageMaker`, `CustomMlModel`).
</ParamField>

<ParamField body="connection" type="object" required>
  Connection configuration specific to the service type.

  <Expandable title="properties">
    <ParamField body="config" type="object">
      Service-specific connection configuration (e.g., `trackingUri`, `registryUri` for Mlflow).
    </ParamField>
  </Expandable>
</ParamField>

<ParamField body="displayName" type="string">
  Human-readable display name for the ML model service.
</ParamField>

<ParamField body="description" type="string">
  Description of the ML model service in Markdown format.
</ParamField>

<ParamField body="owners" type="array">
  Array of owner references (users or teams) to assign to the service.

  <Expandable title="properties">
    <ParamField body="id" type="string">
      UUID of the owner entity.
    </ParamField>

    <ParamField body="type" type="string">
      Type of owner entity (e.g., `user`, `team`).
    </ParamField>

    <ParamField body="name" type="string">
      Name of the owner entity.
    </ParamField>
  </Expandable>
</ParamField>

<ParamField body="domain" type="string">
  Fully qualified name of the domain to assign for governance purposes.
</ParamField>

<ParamField body="tags" type="array">
  Array of classification tags to apply to the ML model service.

  <Expandable title="properties">
    <ParamField body="tagFQN" type="string" required>
      Fully qualified name of the tag.
    </ParamField>

    <ParamField body="labelType" type="string">
      Type of label (e.g., `Manual`, `Derived`, `Propagated`).
    </ParamField>

    <ParamField body="state" type="string">
      State of the tag (e.g., `Suggested`, `Confirmed`).
    </ParamField>
  </Expandable>
</ParamField>

<RequestExample dropdown>
  ```python POST /v1/services/mlmodelServices theme={null}
  from metadata.sdk import configure
  from metadata.sdk.entities import MlModelServices
  from metadata.generated.schema.api.services.createMlModelService import CreateMlModelServiceRequest

  configure(
      host="https://your-company.open-metadata.org/api",
      jwt_token="your-jwt-token"
  )

  request = CreateMlModelServiceRequest(
      name="mlflow_svc",
      displayName="MLflow Production",
      serviceType="Mlflow",
      description="Production MLflow tracking server",
      connection={
          "config": {
              "type": "Mlflow",
              "trackingUri": "http://localhost:8088",
              "registryUri": "http://localhost:8088",
              "supportsMetadataExtraction": True
          }
      }
  )

  service = MlModelServices.create(request)
  print(f"Created: {service.fullyQualifiedName}")
  ```

  ```java POST /v1/services/mlmodelServices theme={null}
  import static org.openmetadata.sdk.fluent.MlModelServices.*;

  MlModelService service = MlModelServices.builder()
      .name("mlflow_svc")
      .serviceType("Mlflow")
      .connection(mlflowConnection()
          .trackingUri("http://localhost:8088")
          .registryUri("http://localhost:8088"))
      .create();
  ```

  ```bash POST /v1/services/mlmodelServices theme={null}
  curl -X POST "{base_url}/api/v1/services/mlmodelServices" \
    -H "Authorization: Bearer {access_token}" \
    -H "Content-Type: application/json" \
    -d '{
      "name": "mlflow_svc",
      "serviceType": "Mlflow",
      "description": "Production MLflow tracking server",
      "connection": {
        "config": {
          "type": "Mlflow",
          "trackingUri": "http://localhost:8088",
          "registryUri": "http://localhost:8088",
          "supportsMetadataExtraction": true
        }
      }
    }'
  ```
</RequestExample>

<ResponseExample>
  ```json Response theme={null}
  {
    "id": "ca22d46e-81b9-4e48-85b5-0adc44980da9",
    "name": "mlflow_svc",
    "fullyQualifiedName": "mlflow_svc",
    "serviceType": "Mlflow",
    "description": "Production MLflow tracking server",
    "version": 0.1,
    "updatedAt": 1769982621618,
    "updatedBy": "admin",
    "href": "http://localhost:8585/api/v1/services/mlmodelServices/ca22d46e-81b9-4e48-85b5-0adc44980da9",
    "connection": {
      "config": {
        "type": "Mlflow",
        "trackingUri": "http://localhost:8088",
        "registryUri": "http://localhost:8088",
        "supportsMetadataExtraction": true
      }
    },
    "owners": [],
    "tags": [],
    "deleted": false,
    "domains": []
  }
  ```
</ResponseExample>

***

## Returns

Returns the created ML model service object with all specified properties and system-generated fields.

## Response

<ResponseField name="id" type="string">
  Unique identifier for the ML model service (UUID format).
</ResponseField>

<ResponseField name="name" type="string">
  ML model service name.
</ResponseField>

<ResponseField name="fullyQualifiedName" type="string">
  Fully qualified name of the service.
</ResponseField>

<ResponseField name="displayName" type="string">
  Human-readable display name.
</ResponseField>

<ResponseField name="description" type="string">
  Description of the ML model service in Markdown format.
</ResponseField>

<ResponseField name="serviceType" type="string">
  Type of ML model service (e.g., Mlflow, Sklearn, SageMaker, CustomMlModel).
</ResponseField>

<ResponseField name="connection" type="object">
  Connection configuration for the service.

  <Expandable title="properties">
    <ResponseField name="config" type="object">
      Service-specific connection configuration.
    </ResponseField>
  </Expandable>
</ResponseField>

<ResponseField name="owners" type="array" optional>
  List of owners assigned to the ML model service.

  <Expandable title="properties">
    <ResponseField name="id" type="string">
      UUID of the owner entity.
    </ResponseField>

    <ResponseField name="type" type="string">
      Type of owner entity (e.g., `user`, `team`).
    </ResponseField>

    <ResponseField name="name" type="string">
      Name of the owner entity.
    </ResponseField>
  </Expandable>
</ResponseField>

<ResponseField name="domain" type="string" optional>
  Fully qualified name of the assigned domain.
</ResponseField>

<ResponseField name="tags" type="array" optional>
  Classification tags applied to the ML model service.

  <Expandable title="properties">
    <ResponseField name="tagFQN" type="string">
      Fully qualified name of the tag.
    </ResponseField>

    <ResponseField name="labelType" type="string">
      Type of label (e.g., `Manual`, `Derived`, `Propagated`).
    </ResponseField>

    <ResponseField name="state" type="string">
      State of the tag (e.g., `Suggested`, `Confirmed`).
    </ResponseField>
  </Expandable>
</ResponseField>

<ResponseField name="version" type="number">
  Version number for the entity (starts at 0.1).
</ResponseField>

***

## Create or Update (PUT)

Use `PUT /v1/services/mlmodelServices` instead of `POST` to perform an upsert. If an ML model service with the same `fullyQualifiedName` already exists, it will be updated; otherwise, a new service is created. The request body is the same as `POST`.

```bash theme={null}
curl -X PUT "{base_url}/api/v1/services/mlmodelServices" \
  -H "Authorization: Bearer {access_token}" \
  -H "Content-Type: application/json" \
  -d '{ ... same body as POST ... }'
```

<Note>
  `PUT` will not return a `409` conflict error if the entity already exists -- it will update the existing entity instead.
</Note>

***

## Bulk Create or Update (PUT)

Use `PUT /v1/services/mlmodelServices/bulk` to create or update multiple ML model services in a single request. The request body is an array of create request objects.

```bash theme={null}
curl -X PUT "{base_url}/api/v1/services/mlmodelServices/bulk" \
  -H "Authorization: Bearer {access_token}" \
  -H "Content-Type: application/json" \
  -d '[
    { "name": "mlflow_prod", "serviceType": "Mlflow", "connection": { "config": { "type": "Mlflow", "trackingUri": "http://mlflow-prod:8088" } } },
    { "name": "sagemaker_prod", "serviceType": "SageMaker", "connection": { "config": { "type": "SageMaker" } } }
  ]'
```

***

## Error Handling

| Code  | Error Type              | Description                                                |
| ----- | ----------------------- | ---------------------------------------------------------- |
| `400` | `BAD_REQUEST`           | Invalid request body or missing required fields            |
| `401` | `UNAUTHORIZED`          | Invalid or missing authentication token                    |
| `403` | `FORBIDDEN`             | User lacks permission to create ML model services          |
| `409` | `ENTITY_ALREADY_EXISTS` | ML model service with same name already exists (POST only) |