Skip to main content

Manage AI Providers

Adding your own provider points the AI functionality of Digital.ai products at your own vendor account instead of the built-in Digital.ai provider. For more background and context, see Bring Your Own LLM.

This topic covers the full lifecycle of managing AI providers: gathering credentials, adding a provider, testing it, making it the default, updating it, and removing it.

warning

API keys and AWS access keys grant access to your vendor account and any spend associated with it. Treat them as secrets, share them only with people who need them, and rotate them on your organization's normal schedule. For more information, see Update Provider Credentials.

Prerequisites​

Collect the following credentials from your vendor before you start. The exact field names for each type are listed in AI Provider Configuration Reference.

note

Vendor consoles and their wording change often. If the steps at your vendor don't match what you see, check the vendor's current documentation using the links in this section.

OpenAI​

  • API key (required): Create one in the OpenAI platform, on the API keys page under Settings. The key is shown only once, so copy and store it securely. See OpenAI API keys.
  • Organization ID (optional): Found on your organization settings page. Use it to attribute requests to a specific organization when your account belongs to more than one.
  • Project ID (optional): Found on the settings page for the project. Use it to scope requests and billing to one project. See Managing projects in the API platform.
  • Base URL (optional): Needed only if you reach OpenAI through a proxy or gateway rather than the standard endpoint.

Anthropic​

  • API key (required): Sign in to the Anthropic Console, go to the API keys page under Settings, and create a key. Keys begin with sk-ant-, are tied to a workspace, and are shown only once, so copy the key when you create it. See Anthropic Console API keys.
  • Base URL (optional): Needed only if you reach Anthropic through a custom endpoint or proxy.

AWS Bedrock​

  • Region (required): The AWS region to connect to, such as us-east-1. A provider connects to a single region. See Amazon Bedrock quickstart.
  • Access key ID and secret access key (required): For an IAM principal that has Bedrock permissions in your AWS account. Enter both together. See Identity and access management for Amazon Bedrock.
  • Session token (temporary credentials only): Required when you authenticate with short-lived AWS Security Token Service (STS) credentials instead of a long-lived key. A session token requires both the access key ID and the secret access key.
  • Model access: Confirm that the models you plan to use are available to your account in that region. See Request access to models.

For batch operations, you also need an S3 bucket for input and output, and an IAM service role that Bedrock assumes to reach it. See Create a custom service role for batch inference.

Beyond credentials and region, no other field is needed for a typical connection. Everything else is optional.

Add a Provider​

  1. Log in to the Platform as an administrator.
  2. In the left navigation, click Settings then click AI providers.
  3. On the Providers tab, click Add provider.
  4. In Name, enter a unique name for this provider. The name can't start with the reserved dai- prefix, and you can't change it after you create the provider.
  5. In Type, select AWS Bedrock, OpenAI, or Anthropic. The configuration section below rebuilds for the type you choose, and you can't change the type after you create the provider.
  6. Complete the fields in the configuration section. A typical setup needs only the fields shown up front; the rest sit under Additional Configuration Options. Credential fields are masked as you type. For the full list, see AI Provider Configuration Reference.
  7. Optionally, in Description, add a note about the provider.
  8. Optionally, select Set as default provider to route all AI operations to this provider as soon as you create it.
  9. Click Create.
note

This button stays disabled until you've entered a name, chosen a type, and completed the required configuration fields. If a required field is empty, the form shows Please fill in all required fields.

Test a Provider​

Testing confirms that the credentials and settings work before you rely on the provider.

  1. On the Providers tab, find the provider's row.
  2. In the Actions column, click the Test icon.
  3. Wait for the result. A message confirming the configuration is valid means the connection works. A failure message means you should re-check the configuration and try again. See Update Provider Credentials.
note

You can't test the built-in Digital.ai provider. It's managed by Digital.ai and always available.

Set the Default Provider​

The default provider handles all AI operations. Only one provider can be the default at a time.

  1. On the Providers tab, select the Default option on the provider's row.
  2. In the Switch default AI provider dialog, click Switch.

Update Provider Credentials​

Edit a provider to replace an expiring key or change a setting. Name and Type are locked after creation.

  1. On the Providers tab, click the Edit icon on the provider's row.
  2. Update the fields that changed, such as a new API key.
  3. Click Update.
  4. Test the provider to confirm the new credentials work.
note

Update stays disabled until you change the configuration or the description, because the name and type can't be edited.

Delete a Provider​

Deleting a provider can't be undone.

  1. If this provider is the current default, the Digital.ai provider is set as the default.
  2. Repoint any model mappings that use this provider. See Update a Mapping After a Provider Retires a Model.
  3. On the Providers tab, click the Delete icon on the provider's row.
  4. In the Delete AI provider dialog, click Delete.
note

You can't delete the built-in Digital.ai provider.

Move to a Different Provider​

To switch vendors, or move to a different account with the same vendor, add the new provider before you remove the old one so AI operations keep working throughout.

  1. Add the new provider and test it.
  2. Repoint your model mappings to the new provider. See Map an Alias to Your Models.
  3. Set the new provider as the default.
  4. Delete the old provider.
Last updated on