Super Admins can configure custom models for managed use globally across the organization. Group Admins can configure custom models for use by the groups they manage.
Note: This article covers adding a custom model to the platform. To learn how to add a pre-configured model or provider, see Add a Public Model/Provider to the Platform.
Click Model Configurations in the navigation panel.
The Model Configurations screen opens. Click the Global or Groups tab. The list of pre-configured LLM providers appears. Each tile displays one provider. Click Add on a tile to add a model supported by that provider.
Add the Model
From the Global or Groups screen, click Add other providers to include a new provider to the list of available providers. The Other models screen opens.
Note: This feature enables connections to any provider, including providers the CalypsoAI platform does not currently host, as well as self-hosted models through Virtual Large Language Model (vLLM), for example.
Connect to Model Provider
Open the Select config drop-down list and select the provider.
Enter a name for the model in the Model Display Name text box. The Inputs and Configuration windows will display the model parameters and outputs, respectively, for that model.
Secrets
If the provider requires a secret, such as an API key, to be provided via an authentication header, we can supply this securely through the Secret section in the Configuration window and paste it into the Key field. The value of the key will be displayed in the Value field.
Input
Enter changes to the model parameters via the Input window.
Configuration
Enter changes to the model configuration via the Configuration window.
Configuration
Enter changes to the model configuration via the Configuration window.
Finalize Changes
When the preferred changes have been made, click Test and Connect. An error message appears if the test and/or connection was not successful.
Please note that the CalypsoAI team is available for assistance with configuration changes for custom models.