Skip to main content
The Model Library is a central inventory of every AI model available to your organization. It is the single, reliable source of the model connection information other modules — Guard, Sandbox, Agent Builder, and Secondary Model Routing — depend on.

​
Registering models

Add a model with the Add Model action, defining:
  • Name — a display name.
  • Provider — for example local or openai.
  • Model identifier — the provider’s model ID.
  • Base URL — the endpoint the model is accessed through.
The list combines local / self-hosted models and cloud models (such as OpenAI) under one view, giving full visibility into which models are registered, by which organization, and when.

​
Credentials and Fetch Models

On a model’s edit screen, the authentication credentials needed to connect (Header Key / Header Value) are stored securely and displayed masked by default. The Fetch Models feature queries the provider using the entered connection details and returns the models currently available, so you can select and update from the live list.
Because model connections live in one place, updating a model’s details once propagates automatically to every module that uses it — there’s no need to maintain separate integrations per module.

​
The foundation for other modules

The Model Library is the reference data layer the rest of the platform builds on:

Sandbox

Pulls models from the registry to security-scan them.

Secondary Model Routing

Uses the primary–secondary mapping defined here.

Proxies & Endpoints

Points endpoints at registered models.

Red Teaming

Tests registered models against attack scenarios.