A file in the document (type "attachment"), for example an image.

The JSON keys are camelCase, as in Lexical: key, url, name, contentType, bytes, width, height, preview and caption.

In HTML, an attachment is a figure with the class kotoba-attachment. Its caption, when it has one, is its text; else its file name:

• An image has an img, with the caption (else the name) as its alt, and a figcaption with the caption (else the name).
• A PDF (figure.kotoba-attachment-pdf) has an object that shows it in the browser's viewer, with a download link inside for a browser that has none; with a preview, a link to the PDF with the preview image instead.
• A video (figure.kotoba-attachment-video) has a video with controls, preload="metadata", no autoplay, the preview as its poster, and a download link inside.
• Another file has only its figcaption. The figcaption of a PDF, a video and another file is a download link, with the caption (else the name) as its text; the caption also names the viewer and the player (aria-label).

As plain text, an attachment is its caption, else its name; in Markdown, ![caption](url) for an image and [caption](url) for another file, with the name when there is no caption. With the :untrusted policy, only the caption (else the name) is rendered, escaped.

• key - the storage key of the file.
• url - the URL of the file.
• name - the file name.
• content_type - the media type, for example "image/png".
• bytes - the size of the file.
• width and height - the size of an image in pixels, or nil.
• preview - the URL of an image of a PDF's first page or a video's poster frame, or nil. The app makes it (see the :preview option of Kotoba.Live.consume_uploads/4).
• caption - what the attachment shows, written by the person, or nil: for an image, its alternative text and caption. The editor writes it under the selected attachment; Kotoba.Content.from_markdown/2 reads it from the alternative text of ![caption](attachment:name).