Glossary

Assistant: a purpose-built helper with its own persona, knowledge, and (optionally) tools.

Chatbot: a plain, governed model chat bound to one model, with no custom knowledge or tools.

System assistant: a built-in helper that powers the platform (Receptionist, Document Analyst, and the AI drafters).

Receptionist: the system assistant that routes people to the right assistant from a plain-language description.

Capability: a named permission that gates both what a user sees and what the API allows.

Role: a named set of capabilities assigned to users. Can be a built-in system role or a custom one.

Team: a group of users, used to share assistants, documents, and chats.

Tier: a subscription level (Starter, Professional, Enterprise) that determines which features are available.

Deployment: a specific model made available in an environment; assistants run on one, often chosen by tier.

Knowledge base: the documents an assistant can draw on for grounded, cited answers.

Grounding / RAG: answering from retrieved passages of your documents rather than the model’s memory alone. See how grounding works.

Guardrails: input/output checks every conversation passes through, including PII detection.

Draft / version: pending edits (draft) versus a published, immutable snapshot (version). See drafts and versions.

Connection: a reusable link to an external system; tools are the actions an assistant takes through it. See tools vs connectors.

Prompt / task / form: a saved prompt offered as a scheduled task, an API, or a fillable form.

Module: a bundle of assistants distributed to an environment centrally.