Skip to main content
Agent: A system that uses tools and multiple steps to pursue a defined task. API: An application programming interface that lets software systems communicate. Audit event: A record of a relevant action, access decision, or security event. Checkpoint: A saved model or training state that can be restored under documented conditions. Connector: A controlled integration between Orbit and an external service. Evaluation: A structured test used to measure a model or system against defined criteria. Inference: The process of running a trained model to produce an output. Model router: A component that selects an available model based on policy and request requirements. Multimodal: Able to process or generate more than one type of data, such as text and images, when implemented. Pulsar: The intended name for Orbit’s model research and development family. Prompt injection: An attempt to manipulate an AI system through untrusted content or instructions embedded in data. RBAC: Role-based access control, where permissions are assigned through defined roles. Sandbox: An isolated environment that limits what code or a process can access. Safety case: A structured argument supported by evidence that a system is acceptably safe for a defined use and environment. Tool calling: A model-driven request for an external function or service to perform a specific operation. Workspace: The user-facing environment for conversations, projects, files, and supported AI tasks.