Skip to main content

Configure a new model

Add a model configuration so you can use that model in chats, agents and swarms.

Before you start​

  • To use a provider with your own key, have your API key for that provider ready.
  • To use a self-hosted model, have your endpoint's base URL, the model identifier your endpoint expects and, if needed, its API key.

Add a model from a provider with your own API key​

  1. On the Models screen, select Configure New Model. The Model Configuration Builder page opens.
  2. In Source, select External API key.
  3. In Provider, select the provider.
  4. In Select Model, tick every model you want to configure. Select Select all to tick them all. One configuration is created for each model.
  5. In API Key, paste your API key for the provider. The key is checked with the provider when you save, and stored encrypted.
  6. Check the name in Model Name. It is filled in for you, and it is shown in the model list and the chat model dropdown. If you ticked several models, you can check each name under Model Names. If you clear a name, the model's catalogue name is used.
  7. Under Visibility, select Org so anyone in your account can use the model, or Private so only you can see and use it.
  8. Select Create Configuration.

Add a self-hosted model​

  1. On the Models screen, select Configure New Model.
  2. In Source, select Self Hosted.
  3. In Provider, select the provider.
  4. In Model Name, enter the identifier your endpoint expects. This name is also shown in the model list and the chat model dropdown.
  5. In Base URL, enter the root URL of your endpoint, including http:// or https://.
  6. In API Key, paste the key your endpoint requires. Leave it blank if your endpoint needs no key.
  7. Under Visibility, select Org or Private.
  8. Select Create Configuration.

Select Cancel or Back to Model Configuration to leave without saving.

What happens next​

A message confirms the model has been configured, and the model appears in the list.

If you ticked several models, a progress card on the Models screen shows how many are configured and how many were skipped. Stay on the screen until it finishes. Leaving the screen stops the rest.

Troubleshooting​

If the form shows Fix the highlighted fields and try again., correct the fields marked in red:

MessageFix
Please select a sourceChoose a Source.
Please select a providerChoose a Provider.
Please select a modelTick at least one model.
Please enter a model nameEnter a Model Name.
Model name must be 200 characters or fewerShorten the name.
Enter your Provider API key.Paste your API key.
Enter the base URL of your endpoint.Enter a Base URL.
Base URL must use http:// or https://.Add the scheme to the start of the URL.
Enter a URL that includes a host name.Check the URL is complete.
This model is already configured. Open the Org Models tab to edit the existing configuration.Edit the existing configuration instead.
This model is no longer available from its provider and has been removed from the list. Choose another model.Pick a different model.

If the name is already used by another configuration, choose a different name.