@webority/markdown-core

The one markdown grammar every Webority renderer reads. It turns a markdown string into blocks and inline tokens, decides what a link may open, and extracts plain text. It draws nothing and touches no platform API, so AppMarkdown (React), <wui-markdown> and <app-markdown> (Razor pages) and the phone Markdown all draw the same document from the same rules.

| Export | What it does | | --- | --- | | parseMarkdown(source) | the blocks: paragraph, heading, list (nested by depth), table, code, quote, rule | | parseInline(text, options?) | the inline tokens: text, code, strong, em, strike, link, image, citation | | splitCitations(text, isCitation) | the plain text split at its [inner] runs: { text } parts and { citation } parts, a run isCitation rejects stays text | | CITATION_RUN | the one citation pattern: [inner] not followed by (, inner non-empty with no bracket or line break | | linkTarget(href, links, resolveLink, origin?, base?) | what a link opens: an app route, a page of this site (a relative address, or a web address on origin), a web address, a mail or phone address, or nothing. With base (an origin from linkBase) a relative address resolves against it first | | linkBase(value) | the origin a relative link resolves against: an https origin with no path, query, fragment or credentials (a trailing / allowed), without it; anything else is null | | webRoute(target) | the web route of a resolved app link | | markdownToPlainText(source) | the text a screen reader, a preview or a search index reads | | FigureRun | the type a figureText splitter returns: { kind: "text" \| "sub" \| "sup", text }; a sub or sup run draws lowered or raised |

links is "web" (open web, mail and phone addresses), "app" (open only what resolveLink resolves to an in-app route) or "page". In "page" mode resolveLink is asked about every address, and a relative address or a web address on origin draws as a page link; other web and contact addresses stay text. Chat replies use "app", because model output must not send a person off the app.

Citations are chat-only. parseInline(text, { isCitation }) reads an [inner] run that is not a link as a citation token ({ type: "citation", text: inner }) when isCitation(inner) returns true; without isCitation, no run is a citation. Citations are read inside bold, italics and strike, never inside a link label or image alt text, and an escaped \[ is never a citation. splitCitations applies the same rule to plain text for renderers that do not parse inline tokens.