Renders model responses as plain text or streamed Markdown.

Response is intentionally small: it handles the response surface and, when format="markdown", uses MDEx to render sanitized Markdown. It does not own chat state, provider callbacks, or streaming transport. The parent LiveView passes the current text as it changes.

Examples

<.response value={@answer} />

<.response id="answer" value={@answer} streaming reveal="word" />

<.response value={@streamed_markdown} format="markdown" streaming />

<.response>
  <p>Custom rendered content.</p>
</.response>

Attributes

• value - Response text or Markdown. Defaults to nil.
• format - One of text or markdown. Defaults to text.
• streaming - Marks the response as incomplete. The parent LiveView owns this state.
• reveal - One of chunk, word, character, or line. Defaults to chunk.
• sanitize - Sanitizes Markdown HTML output. Defaults to true.
• mdex_options - Extra MDEx options passed to the renderer.
• class - Additional CSS classes.

Animated reveal modes (word, character, line) require an id because Phoenix LiveView hooks need a stable DOM id. They smooth already-received text in the browser; the parent LiveView still owns provider streaming. Markdown always renders by chunk so partial Markdown can stay valid and sanitized.

Accessibility

• When streaming is true, the root sets aria-live="polite" unless you pass your own aria-live value, and aria-busy="true" unless you pass your own value.
• Markdown output is sanitized by default before being marked as safe HTML.