anydoc

type Asset struct {
	// ID is the document-model asset index, as referenced by in-place image
	// sources. It is stable for a single conversion.
	ID uint64
	// Name is a generated, extension-carrying file name ("image-1.png"):
	// embedded assets have no name of their own inside the container.
	Name string
	// MediaType is the IANA media type the container declared.
	MediaType string
	// Data is the raw image bytes.
	Data []byte
	// Alt is the image's alternative text as the document author wrote it,
	// empty when the document carries none.
	Alt string
	// Section is the text of the nearest heading above the image, empty when
	// the image sits before the first heading.
	Section string
}

type Options struct {
	// Format names the parser explicitly ("docx", "csv", ...). Empty means
	// the format is detected from the content, which every format except
	// CSV carries a signature for.
	Format string
	// WithAssets extracts embedded images and renders them in place in the
	// markdown using anydoc's official GFM serializer (asset images are
	// rewritten to `images/image-N.ext` links first). Ignored for PDF.
	WithAssets bool
}

type Result struct {
	// Markdown is GitHub-Flavored Markdown for the whole document.
	// When WithAssets is set, embedded images appear in place as
	// `![alt](images/image-N.ext)` via anydoc's official serializer.
	Markdown string
	// Assets are the images embedded in the document, in document order.
	// Always empty for PDF, which anydoc renders straight to Markdown
	// without a document model.
	Assets []Asset
	// AssetsError explains why Assets is empty when images were asked for.
	// The document-model parse is what both extracts images and places them
	// in the markdown; when it fails, Markdown still comes from the text-only
	// renderer so the conversion succeeds and the caller decides whether to
	// log or ignore the loss.
	AssetsError error
}