# MarkdownLoader Loads UTF-8 Markdown into a `MarkdownDocument` syntax tree, without a renderer or UI dependency. ```typescript import {parse} from '@loaders.gl/core'; import {MarkdownLoader} from '@loaders.gl/documents'; const document = await parse('# Title\n\nA **bold** statement.', MarkdownLoader); ``` For synchronous parsing, import `MarkdownLoaderWithParser` from `@loaders.gl/documents/markdown` and use `parseSync`. | Property | Value | | --- | --- | | Extensions | `.md`, `.markdown` | | MIME type | `text/markdown` | | Input | UTF-8 bytes or text | | Output | `MarkdownDocument` | | Worker | No | ## Options `options.markdown.maxDepth` bounds nested quotes and inline markup. It defaults to `32` and must be an integer from `1` to `128`, including the root level. Exceeding the limit throws an error. ## Output The root is `{type: 'root', children: MarkdownBlock[]}`. Blocks include headings (`depth` and inline `children`), paragraphs (`children`), fenced code (`value` and optional `language`), flat lists (`ordered`, optional numeric `start`, and inline `items`), blockquotes (`children`), and thematic breaks. Inline nodes include text and code (`value`), emphasis and strong text (`children`), links (`url`, `children`), and images (`url`, plain `alt`). Types are exported from the package root and `/markdown-types`. ## Supported subset Supports ATX headings, paragraphs, backtick/tilde fences, blockquotes, single-line ordered and unordered list items, thematic breaks, inline code, emphasis, strong text, inline links/images, and backslash punctuation escapes. Soft line breaks remain newline characters in text nodes. Unclosed fences consume the remaining document; unmatched inline delimiters remain literal text. This is a deliberately small parser, not a CommonMark or mdast implementation. Nested/multiline lists, setext headings, indented code, reference links, tables, task lists, autolinks, full delimiter precedence, HTML parsing, and MDX are not supported. Choose a fuller Markdown engine when those semantics are required. Raw HTML remains text. Destinations remain authored strings: the parser never fetches links or images, executes code, or renders HTML. Applications rendering the tree must escape text and validate URL schemes according to their own policy.