Create Provider Account

Use /app/form/createProviderAccount to create the non-secret provider billing and quota identity that an LLM Deployment will use. This is not a Portal user account and it does not store an API key, access token, password, or provider endpoint.

Open the form from Administration > GenAI Admin > LLM Models > Accounts by choosing Create provider account.

Before You Begin

Create an Account before creating a Deployment that refers to it. Decide which provider, billing owner, and quota pool the new Account represents. If one host uses multiple provider subscriptions, projects, or cost centers, create a separate Account for each boundary that must be managed independently.

Form Fields

The backend creates providerAccountId. The form does not ask you to supply that identifier. The active field is also backend-managed and is not part of the form.

Billing Principal

Use billingPrincipal to identify who is financially responsible for usage. The exact value depends on your organization and provider. Examples include:

• genai-platform-cost-center
• azure-subscription-production
• aws-account-llm-platform
• provider-project-customer-support

This field is governance and audit metadata. Do not enter a provider API key, secret value, bearer token, password, or authorization header.

Quota Group Id

Use quotaGroupId to name the provider capacity pool shared by deployments, for example openai-production-capacity. Keep the value stable and use the same intended quota-group identity when configuring the related Deployment and gateway publication.

Deployments in the same published quota group share the corresponding gateway capacity identity. Changing an Account later does not automatically rewrite an already published gateway snapshot.

Capacity Metadata

capacityMetadata is an optional open-ended object for non-secret capacity annotations. The editor supports JSON and YAML. For example:

{
  "serviceTier": "production",
  "approvedRpm": 1000,
  "approvedTpm": 2000000
}

The equivalent YAML is:

serviceTier: production
approvedRpm: 1000
approvedTpm: 2000000

Use an empty object when no metadata is needed:

{}

After editing JSON or YAML, choose Apply. The Create action remains blocked while the structured draft is invalid or has unapplied changes. Capacity metadata is currently retained for control-plane governance; the gateway does not use arbitrary metadata properties to resolve credentials or select routes.

Create the Account

Review the values and choose Create Provider Account. The form sends the lightapi.net/genai/createLlmProviderAccount/0.1.0 command. After a successful command, the browser returns to Administration > GenAI Admin > LLM Models. The new Account can then be selected by a provider Deployment.

NVIDIA free embedding demo

For the NVIDIA hosted nemotron-3-embed-1b demo, use:

{
  "accountName": "nvidia-free-embedding-demo",
  "providerType": "nvidia",
  "billingPrincipal": "nvidia-build-api-demo",
  "quotaGroupId": "nvidia-free-embedding-demo",
  "capacityMetadata": {
    "serviceTier": "free-demo",
    "sharedExternalQuota": true
  }
}

billingPrincipal and quotaGroupId are operator-assigned governance names; they are not the NVIDIA API key. If your NVIDIA account exposes a stable organization or project identifier, use that identifier for Billing Principal instead. Keep all secret material on the Credentials tab.

Common Problems

• Provider Type is empty: confirm the model_provider reference values are configured and available to the Portal environment.
• Required-field validation: provide Account Name, Provider Type, Billing Principal, and Quota Group Id.
• Capacity Metadata error: enter an object rather than an array or quoted JSON string, correct any JSON/YAML syntax error, and choose Apply.
• Duplicate account: choose a different Account Name. A host cannot contain two Accounts with the same Provider Type and Account Name.
• 403 on Create: confirm access to the lightapi.net/genai/createLlmProviderAccount/0.1.0 command endpoint and the required write permission.

For account relationships and gateway usage, see the LLM Model Control Plane guide.