Zotero

Zotero

Use this skill to operate a user's local Zotero Desktop library.

Connection Setup

Before any operation, ensure Zotero is running and the local API is enabled:

1. Open Zotero Desktop application
2. Enable Local API: Edit > Preferences > Advanced > check "Allow other applications on this computer to communicate with Zotero" (可用于 http://localhost:23119/api/)
3. Verify connection:

PowerShell (Windows)

powershellCopy code
Invoke-WebRequest -Uri "http://127.0.0.1:23119/api/users/0/collections" -Headers @{ "Zotero-API-Version" = "3" } -UseBasicParsing

curl (cross-platform)

bashCopy code
curl -s -H "Zotero-API-Version: 3" "http://127.0.0.1:23119/api/users/0/collections"

If the connection fails with "connection closed":

• The local API preference is not yet enabled
• Close and reopen Zotero after checking the preference
• Or manually add to prefs.js: user_pref("extensions.zotero.httpServer.localAPI.enabled", true);

Direct HTTP API Usage

When the Python helper script is unavailable or fails, use direct HTTP calls to the Zotero local API (port 23119).

List collections (folders)

bashCopy code
curl -s -H "Zotero-API-Version: 3" "http://127.0.0.1:23119/api/users/0/collections"

Response includes key, name, and meta.numItems for each collection.

List items in a collection

bashCopy code
# Top-level items only
curl -s -H "Zotero-API-Version: 3" "http://127.0.0.1:23119/api/users/0/collections/{collectionKey}/items/top?limit=25"

# All items (including attachments)
curl -s -H "Zotero-API-Version: 3" "http://127.0.0.1:23119/api/users/0/collections/{collectionKey}/items?limit=25"

Search across library

bashCopy code
curl -s -H "Zotero-API-Version: 3" "http://127.0.0.1:23119/api/users/0/items?q=transformer&limit=10"

Export BibTeX

bashCopy code
curl -s -H "Zotero-API-Version: 3" "http://127.0.0.1:23119/api/users/0/items?format=bibtex&limit=100"

Get item children (attachments)

bashCopy code
curl -s -H "Zotero-API-Version: 3" "http://127.0.0.1:23119/api/users/0/items/{itemKey}/children"

Get attachment file URL

bashCopy code
curl -s -H "Zotero-API-Version: 3" "http://127.0.0.1:23119/api/users/0/items/{attachmentKey}/file/view/url"

Reading PDFs from Zotero

Zotero stores PDFs in its local storage. To read PDF content:

1. Find the PDF local path

Zotero stores attachments under %USERPROFILE%\Zotero\storage\{attachmentKey}\.

Get the attachment key from the item's children:

bashCopy code
curl -s -H "Zotero-API-Version: 3" "http://127.0.0.1:23119/api/users/0/items/{itemKey}/children"

The response includes links.attachment.href and data.filename.

2. Locate the PDF file

The PDF is typically at:

Copy code
%USERPROFILE%\Zotero\storage\{attachmentKey}\{filename}.pdf

Example:

powershellCopy code
Get-ChildItem "$env:USERPROFILE\Zotero\storage\HQV9AFD4"

3. Read PDF content with pdfplumber

Install pdfplumber (requires Python):

bashCopy code
py -m pip install pdfplumber

Extract text:

pythonCopy code
import pdfplumber
pdf_path = r"C:\Users\{user}\Zotero\storage\{attachmentKey}\{filename}.pdf"
with pdfplumber.open(pdf_path) as pdf:
    for i, page in enumerate(pdf.pages[:5]):  # first 5 pages
        text = page.extract_text()
        print(f"--- Page {i+1} ---")
        print(text[:2000] if text else "(no text)")

4. Alternative: Use Codex bundled Python

If system Python lacks pip, use Codex's bundled Python:

powershellCopy code
$python = "C:\Users\{user}\.cache\codex-runtimes\codex-primary-runtime\dependencies\python\python.exe"
& $python -m pip install pdfplumber
& $python -c "import pdfplumber; ..."

Python Helper Script

When available, the stdlib-only helper script provides convenience:

bashCopy code
py <skill-root>/scripts/zotero.py <command>

Note for macOS/Linux users: Use python3 instead of py if the py launcher is not available.

Fast starts

Check readiness:

bashCopy code
py scripts/zotero.py status --json

Enable local API and restart Zotero:

bashCopy code
py scripts/zotero.py enable --restart

Search and export:

bashCopy code
py scripts/zotero.py search "transformer" --json
py scripts/zotero.py export-bibtex --out references.bib

Workflow

1. Verify Zotero is running and local API is enabled.
2. Choose connection method:
   • Prefer direct HTTP API calls when the Python helper is unavailable
   • Use the helper script when available for convenience
3. Read-only operations (safe by default):
   • List collections/items via HTTP API
   • Search library
   • Export BibTeX
   • Read PDFs from local storage
4. Attachment/PDF access: Only retrieve when explicitly requested
5. Write operations: Confirm with user before import/save actions

Common HTTP API Patterns

Summarize a collection

bashCopy code
# 1. Get collection key
curl -s -H "Zotero-API-Version: 3" "http://127.0.0.1:23119/api/users/0/collections"

# 2. Get items in collection (extract title, abstract, creators)
curl -s -H "Zotero-API-Version: 3" "http://127.0.0.1:23119/api/users/0/collections/{key}/items/top?limit=50"

# 3. For each item, get children to find PDF attachment key
curl -s -H "Zotero-API-Version: 3" "http://127.0.0.1:23119/api/users/0/items/{itemKey}/children"

# 4. Read PDF from %USERPROFILE%\Zotero\storage\{attachmentKey}\

Export collection to BibTeX

bashCopy code
curl -s -H "Zotero-API-Version: 3" "http://127.0.0.1:23119/api/users/0/collections/{key}/items?format=bibtex" > collection.bib

Output standards

• For inventory/search, return title, creators, year, Zotero item key, and BibTeX key when available.
• For PDF reading, summarize key findings: research question, method, results, and conclusions.
• For .bib export, return the absolute output path and entry count.
• For blockers, name the exact gate: Zotero app missing, local API disabled, port closed, no matching item, or PDF not found.

Route details

Read references/local-api-routes.md for complete endpoint documentation.