Custom resolvers for PDF retrieval currently can be configured via the extensions.zotero.findPDFs.resolvers hidden pref (Advanced preferences → Config Editor). The value should be a JSON string containing an array of configuration objects. HTML and JSON sources are supported.

You may wish to distribute a custom resolver using a Zotero plugin that adds or updates the pref at startup.

{
    "name": "My Paper Source",
    "method": "GET",
    "url": "https://example.com/{doi}",
    "mode": "html",
    "selector": "#pdf-link",
    "attribute": "href",
    "automatic": false
}

selector is a CSS-style selector. If attribute is omitted, the element's textContent is used. index can be passed to select a specific element of a set.

If automatic is false or unspecified, the resolver will be used only for manual actions — Add Item by Identifier and Find Available PDFs — and not when saving from the browser.

Note: The Unpaywall API is used here for demonstration purposes, but Zotero already uses its own mirror of Unpaywall data, so configuring it with a custom resolver isn't necessary.

{
    "name": "Unpaywall",
    "method": "GET",
    "url": "https://api.unpaywall.org/v2/{doi}?email=me@example.com",
    "mode": "json",
    "selector": ".oa_locations.url_for_pdf",
    "automatic": true
}

selector uses JSPath syntax.

In this case, .oa_locations.url_for_pdf matches zero or more URL strings in the url_for_pdf property in objects in the oa_locations array.

However, Unpaywall can return either direct PDF URLs or landing page URLs, so matching on url_for_pdf alone isn't sufficient. For advanced cases like this, a mappings object can be provided to pull out specific values:

{
    "name": "Unpaywall",
    "method": "GET",
    "url": "https://api.unpaywall.org/v2/{doi}?email=me@example.com",
    "mode": "json",
    "selector": ".oa_locations",
    "mappings": {
        "url": "url_for_pdf",
        "pageURL": "url_for_landing_page"
    },
    "automatic": true
}

Here, the oa_locations array is matched directly, and mappings is used to assign the url_for_pdf and url_for_landing_page properties from the array objects to url (for a direct PDF link) and pageURL (for a page where a PDF can be translated).