Defines the Agent’s avatar, including its visual appearance and voice.

Configuration for the Large Language Model used by the Agent to generate responses.
D-ID and Google providers are only supported with Expressive Avatar presenters.

Knowledge Base used by the Agent to retrieve answers (RAG).
Use the ID returned by the Create Knowledge endpoint.

List of suggested questions shown to the user at the start of the chat.

Greeting messages used when the chat starts.
One greeting is selected at random.

Event-based trigger configuration for the Agent.

Images, videos, and gifs the agent can show or reference during a conversation. Each
entry is either an update to an existing asset (provide its id), or a new item added
from your media library (provide media_id, optionally with tags) — its file, name,
description, and thumbnail come from the media library automatically. New assets can
only be added from the media library. Leave out an asset's id or media_id to remove
it from the agent.

Pronunciation dictionary applied to the agent's speech, so matching
words and phrases are spoken with your defined pronunciations. Use null for no dictionary.