Skip to main content

Chat surface

The transcript: messages, streaming markdown, and the indicator shown before the first token arrives. <Chat> composes these for you; reach for them directly when you are building your own layout around a thread.

Streaming answer​

Markdown streams in token by token as the agent replies. ChatMessageList picks StreamingTextRenderer for assistant text and keeps partial markdown (an open code fence, half a list) stable while it grows.

Loading the live demo…

ChatMessageList props​

PropTypeNotes
messagesDistriChatMessage[]Required. From useChat or the chat store.
rendering'rich' | 'minimal'Expanded tool cards, or compact one-line rows.
toolRenderersToolRendererMapYour own renderer per tool name. See Human-in-the-loop.
threadIdstringEnables per-message feedback and trace links.
enableFeedbackbooleanLet the user vote on assistant messages.
onShowTrace(threadId: string) => voidCalled when the user opens a trace.
verbose, debugbooleanMore detail in the transcript, for development.

StreamingTextRenderer props​

PropTypeNotes
textstringRequired. The markdown received so far.
isStreamingbooleanShows the caret and tolerates unfinished markdown.
classNamestring

Thinking indicator​

The shimmering placeholder shown while an agent is working, before the first token lands.

Loading the live demo…

ThinkingRenderer props​

PropTypeNotes
indicatorStreamingIndicatorRequired. 'typing', 'thinking' or 'generating'.
namestringThe agent's display name.
avatarReactNodeShown next to the indicator.
thoughtTextstringReasoning text to show under the indicator, if the model streams it.
classNamestring