A Metalsmith Sanity Source plugin

  • Yogi Berra

In the previous two Ruhrpott blog posts, we have discussed how to build modular page sections with structured content. Initially, we used a Frontmatter YAML object to define a media component, and then we created the same component in Sanity Studio. This blog post will review how to get the Sanity content into the Metalsmith build process.

To quickly review, the Metalsmith build process consists of three stages:

  • Read files from a source folder and transform them into javascript objects
  • Transform objects with plugins
  • Write files to the destination folder

We can substitute the initial Read with fetching page content from Sanity and injecting it into the build process. Since Sanity uses a JSON object to define structured content and the Metalsmith file object is JSON, this process is relatively simple. For our example, there are only a few things to consider.

Rich text fields in Sanity use the Portable Text format , while Metalsmith traditionally uses Markdown.

Images in a Sanity JSON object are referenced with an asset ID; meta properties are not included.

To fetch content from Sanity and transform it into a Metalsmith compatible JSON object we'll build a custom plugin. The plugin uses the Sanity Client Javascript library and a few Sanity utility libraries Block Content to Markdown and Image URL. Let's have a look at the code outline.

The Metalsmith Sanity Source Plugin

function initSanitySource(options) {
  options = normalizeOptions(options);

  return async function metalsmithSourceSanity(files, metalsmith, done) {
    // initialize Sanity client
    const client = sanityClient(options);

    let contentTypes;

     // fetch all content types from Sanity
     const contentTypes = await client.fetch(groq`*[discoverable == true]`

    const data = {};
    contentTypes.forEach((contentType) => {
      // transform Sanity portable text blocks to markdown and resolve image references
      iterate(contentType, client, options);

      if ( contentType.isPage ) {
        // Metalsmith needs contents otherwise it will skip the file
        contentType.contents = Buffer.from('');
        contentType.mode = '0644';
        contentType.stats = {};

        // add page to files object
        const fileKey = `${contentType.slug.current}.md`;
        // add all properties of contentType to files[fileKey]
        files[fileKey] = contentType;

      } else {
        // if content type is not a page, then it is a data file; add to metadata
        // if a contentType._type array is not already in data, add it
        if ( !data[contentType._type] ) {
          data[contentType._type] = [];
          // add this contentType of type contentType._type to the array
        } else {
          // if a contentType._type array is already in data, push the contentType object to its

    // merge data object with existing metadata
    const metadata = metalsmith.metadata();
    metadata['data'] = data;

The plugin fetches all content types with the sanityClient library.

It will then loop over all content types and divide them into page or data objects. This is determined by the boolean contentType.isPage.

If the object represents a page, it adds an object to the files object. If the object represents data, it adds the object to the metadata object.

Metalsmith will do the rest.

Finally, let's have a look at the iterate function. This function transforms Sanity portable text blocks to markdown and resolves image references.

const iterate = (obj, client, options) => {
  Object.keys(obj).forEach(key => {

    // transform Portable Text to Markdown
    if(key === "portableTextBody") {
      obj[key] = BlocksToMarkdown(obj[key], {
        serializers: getSerializers(client),
        projectId: options.projectId,
        dataset: options.dataset,

    // transform image reference to image url
    if(key == "asset" && obj[key]._ref.startsWith("image-")){
      const imageBuilder = imageUrl(client);
      const image = imageBuilder.image(obj[key]);
      obj["imageURL"] = image.url();

    if (typeof obj[key] === 'object' && obj[key] !== null) {
      iterate(obj[key], client, options)

This is where we use the Sanity utility functions blocksToMarkdown and imageUrl .

When we query the Sanity API, rich text content is returned as Portable Text. We transform Portable Text by serializing the arrays that contain Portable Text into Markdown. We are using some custom serializers to do that.

const getSerializers = client => {
  return {
    types: {
      code: ({node}) => '```' + node.language + '\n' + node.code + '\n```',
      mainImage: ({node}) => imageUrl(client).image(node).url(),
      image: ({node}) => `<img src="${imageUrl(client).image(node).url()}" alt="${node.alt}" id="${node.imageId}"/>`,
      slug: ({node}) => node.current,
    marks: {
      link: ({children, mark}) => `<a href="${mark.href}" ${mark.isExternal && "target='_blank' rel='noreferrer, noopener'"}>${children}</a>`,

I found the articles Internal and External Links and Presenting Portable Text helpful in understanding serializers.

The complete code for the Ruhrpott project is available on GitHub. Clone it and experiment, and let me know what you did in the Metalsmith community at Gitter .