sd-export

ScienceDirect Citation Export

Export article citations from ScienceDirect search results. Supports RIS, BibTeX, plain text, and Zotero push.

Export API

ScienceDirect uses a POST endpoint for citation export:

{BASE_URL}/search/api/export-citations

Parameters (form POST):

Field	Description
pii	Article PII (e.g. S0957417426005245). Multiple PIIs for batch.
type	Export format: ris, bibtex, text
t	Security token (must be extracted from the page)

Single Article Export

Step 1: Ensure on search results page

The export API requires a token that exists only on search results pages. If you have a PII but are not on a search results page, first search for the article or navigate back to results.

Note: When navigating to search results, always include initScript to prevent bot detection:

initScript: "Object.defineProperty(navigator, 'webdriver', {get: () => undefined})"

Step 2: Click the article's Export button and extract token

Use evaluate_script:

(pii) => {
  // Find the result item with this PII
  const checkbox = document.getElementById(pii);
  const resultItem = checkbox?.closest('li.ResultItem');
  if (!resultItem) return { error: 'Article with PII ' + pii + ' not found on this page.' };

  // Click the Export button for this result
  const exportBtn = resultItem.querySelector('button[aria-label="Export"]');
  if (exportBtn) exportBtn.click();

  // Wait briefly for the export panel to appear, then extract token
  return new Promise(resolve => {
    setTimeout(() => {
      const form = resultItem.querySelector('.ExportCitationOptions form');
      if (!form) { resolve({ error: 'Export panel did not open.' }); return; }

      const token = form.querySelector('input[name="t"]')?.value || '';
      const formAction = form.action;

      resolve({ token, formAction, pii });
    }, 500);
  });
}

Step 3: Submit export request

Use evaluate_script to submit the form with the desired format:

(token, pii, format) => {
  // format: 'ris', 'bibtex', or 'text'
  const typeMap = {
    'ris': 'ris',
    'bibtex': 'bibtex',
    'text': 'text',
  };

  const form = document.createElement('form');
  form.method = 'POST';
  form.action = window.location.origin + '/search/api/export-citations';
  form.target = '_blank';

  const fields = { type: typeMap[format] || 'ris', t: token, pii: pii };
  for (const [name, value] of Object.entries(fields)) {
    const input = document.createElement('input');
    input.type = 'hidden';
    input.name = name;
    input.value = value;
    form.appendChild(input);
  }

  document.body.appendChild(form);
  form.submit();
  document.body.removeChild(form);

  return { success: true, format: format, pii: pii };
}

Batch Export

Step 1: On search results page, click the top Export button

Use evaluate_script to select multiple articles and trigger batch export:

(piis) => {
  // Select all specified articles
  piis.forEach(pii => {
    const cb = document.getElementById(pii);
    if (cb && !cb.checked) cb.click();
  });

  // Click the batch export button
  const exportBtn = document.querySelector('.export-all-link-button');
  if (exportBtn) exportBtn.click();

  return { success: true, selected: piis.length };
}

Then wait for the export dropdown to appear and select the desired format.

Zotero Push

To push citations to a locally running Zotero instance. Two modes are supported:

Prerequisites: Zotero desktop must be running with the Connector API enabled (default on port 23119).

Mode 1: RIS import (simple, no PDF)

Use when you have RIS data from the export API or constructed from metadata.

python {SKILL_DIR}/scripts/push_to_zotero.py --ris-file "{RIS_FILE_PATH}"

Or push RIS content directly:

python {SKILL_DIR}/scripts/push_to_zotero.py --ris-data "{RIS_CONTENT}"

The script uses a deterministic session ID (MD5 hash of content) so:

• First call → 201 (saved successfully)
• Repeat call → 409 → treated as success ("already saved, no duplicates")

This avoids the SESSION_EXISTS error that occurs with random session IDs, since Zotero's session cleanup is buggy and sessions persist until restart.

Mode 2: JSON import (structured data with optional PDF attachment)

Use when you have structured paper data (e.g., from sd-paper-detail) and want to attach PDFs.

Save paper data as a JSON file, then run:

python {SKILL_DIR}/scripts/push_to_zotero.py --json "{JSON_FILE_PATH}"

JSON format (single paper or array):

{
  "title": "Paper Title",
  "authors": ["Author One", "Author Two"],
  "journal": "Journal Name",
  "date": "2026",
  "doi": "10.1016/...",
  "volume": "76",
  "issue": "3",
  "pages": "109460",
  "abstract": "...",
  "keywords": ["keyword1", "keyword2"],
  "url": "https://www.sciencedirect.com/science/article/pii/{PII}",
  "pdfUrl": "https://www.sciencedirect.com/..../pdfft?md5=...&pid=...",
  "cookies": "cf_clearance=...; JSESSIONID=..."
}

When pdfUrl and cookies are provided, the script will:

1. Save metadata via /connector/saveItems
2. Download PDF using the browser cookies
3. Upload PDF as attachment via /connector/saveAttachment

Listing Zotero collections

python {SKILL_DIR}/scripts/push_to_zotero.py --list

Export Format Buttons on Page

Each result item has these export option buttons (inside .export-options):

Button text	data-aa-button attribute	Format
Save to RefWorks	srp-export-single-refworks	RefWorks
Export citation to RIS	srp-export-single-ris	RIS
Export citation to BibTeX	srp-export-single-bibtex	BibTeX
Export citation to text	srp-export-single-text	Plain text

Notes

• Authentication required: Export buttons, result checkboxes, and the batch action toolbar are only visible to authenticated users. The user must be logged in (institutional or personal account) for export to work.
• The security token t is session-specific and page-specific. It must be extracted from the export form on the page.
• Batch export is much more efficient than exporting one-by-one.
• For Zotero push, ensure Zotero desktop is running before invoking.
• If the page shows captcha during export, auto-click the Cloudflare Turnstile checkbox: use take_snapshot to find checkbox "确认您是真人" inside the Turnstile iframe, then click(uid). Wait 3-5s and verify. If auto-click fails after 2 attempts, fall back to asking the user. After solving once, the session cookie persists.