Paper Library Checker

English | 简体中文

A local-first Zotero companion that tells you whether the paper you are viewing is already saved—before you create a duplicate item.

Paper Library Checker compares metadata from supported scholarly pages, including CNKI, with your local Zotero 9 library and displays Saved, Possible match, or Not saved. Standard mode works with Zotero's built-in Local API and needs no Paper Library Checker Zotero add-on. Enhanced mode adds faster batches, complete fuzzy matching, Possible match, and real-time index updates. The project has no telemetry and no Zotero library upload.

Install from Chrome Web Store · Download releases · Quick start · Supported sites · Privacy · 简体中文

Alpha software: site coverage and installation details may change before the first stable release.

Install from Chrome Web Store — the published store version is 0.4.1.

GitHub Release version is 0.5.0. The published Chrome Web Store listing remains 0.4.1 until its separate store update is submitted and published.

Zotero is a registered trademark of the Corporation for Digital Scholarship. This independent project is not affiliated with or endorsed by the Zotero project.

Why use it?

• Check for an existing Zotero item before clicking Zotero Connector to save.
• Review CNKI articles, search results, and supported reference lists without repeatedly searching Zotero.
• Distinguish confirmed matches from fuzzy matches that need manual review.
• Keep library matching on your computer.

Requirements and compatibility

Component	Current support
Zotero desktop	Zotero 9.0.x only; 9.0.6 is the exact release-tested version.
Google Chrome	The Chrome Web Store is the primary public browser-extension channel. Store version 0.4.1 passed installation, connection, popup, and a representative page check.
Microsoft Edge	Can install the extension from the Chrome Web Store; site support remains limited to the matrix below.
Distribution	Install the browser extension from the Chrome Web Store. The Zotero XPI from GitHub Releases is optional and enables enhanced mode. The GitHub browser ZIP remains available for development, auditing, or manual installation.

The browser extension and Zotero desktop are required. The companion Zotero add-on is required only for enhanced mode. Keep Zotero running while using either mode.

Quick start

1. Install the browser extension from the Chrome Web Store.
2. Start Zotero 9 and enable its built-in Local API if it is disabled.
3. Open the browser extension Options, leave Connection method set to Automatic (recommended) or choose Standard mode, then select Test connection. Standard mode needs no pairing token or XPI.
4. Open a supported article page, select the toolbar icon, and choose Check this page.

Optional enhanced mode

1. Download the Zotero XPI from the GitHub v0.5.0 Release. In Zotero, open Tools > Plugins, choose Install Plugin From File, and install the XPI.
2. Restart Zotero and choose Tools > Paper Library Checker: Copy pairing token.
3. In extension Options, choose Enhanced mode (or keep Automatic), expand the enhanced connection settings, paste the 64-character Pairing token, and select Test connection.

Enhanced mode provides faster reference-list batches, full fuzzy-title matching, Possible match, and real-time index updates.

Mode capability matrix

Capability	Standard / Indexed	Standard / Direct fallback	Enhanced / XPI
XPI required	No	No	Yes
Personal and group libraries	Complete snapshot	Best-effort search	Live add-on index
Complete exact identifier/title recall	Yes, while ready	No	Yes
Complete Not saved result	Yes, while ready	No	Yes
Reference-list batch	Yes	Up to 10 fallback candidates	Yes
Fuzzy title / Possible match	No	No	Yes
Real-time Zotero updates	No	No	Yes

Direct and stale-index misses display No match; result may be incomplete rather than Not saved. See the mode capability contract for canonical capability fields and resolver rules.

Manual/developer installation

For development, auditing, or a manual-installation fallback, download the browser ZIP from the GitHub v0.5.0 Release and extract it to a stable directory. Open chrome://extensions or edge://extensions, enable Developer mode, choose Load unpacked, and select the directory that directly contains manifest.json. Click Reload there after replacing files with a newer manual ZIP. Do not move, rename, or delete the unpacked extension directory after loading it.

Migrating from an unpacked installation

Disable or remove the old unpacked extension before installing the Chrome Web Store version. The store version uses the fixed extension ID pmfobjnkoiiplambnbbkfdlfjcbdogon, so its storage and pairing token may not migrate automatically from an unpacked installation. Paste the token again in Options, click Save, and then Test connection. Do not enable the unpacked and store versions at the same time; duplicate content scripts can produce duplicate status badges or conflicting page state.

Supported sites and status

Site or scenario	Metadata path	Article detail	Reference/list batch
CNKI Chinese	Built-in CNKI extractor	Supported and tested	Experimental
CNKI English	Built-in CNKI/generic extraction	Experimental	Experimental
Scopus	Dedicated document-details extractor and publications-search adapter	Experimental; live DOM verified	Experimental; search results only, in Table and List views
Web of Science	Dedicated Full Record extractor and Summary results adapter	Experimental; live DOM verified	Experimental; Summary search results
MDPI	Generic citation metadata	Supported and tested	Not supported
ScienceDirect	Generic metadata plus a site adapter; optional local translation-server	Best effort; live access may be challenged	Best effort
Springer, Wiley, PubMed, arXiv, IEEE, ACM, Taylor & Francis, and DOI.org	Generic metadata; optional local translation-server	Experimental / best effort	Not supported
Synthetic COinS, JSON-LD, citation, and DC fixtures	Generic extractor	Automated regression coverage only	Not supported

A domain appearing in the browser manifest means that the content script is allowed to run there; it is not by itself a claim of live-site support. Unknown sites are never scanned automatically for references. Broad article detection remains experimental and disabled by default.

The extension does not handle PDFs without usable page metadata, save records, modify the Zotero library, or promise coverage equivalent to Zotero Connector translators. Publishing through the Chrome Web Store does not expand the site-support claims in this table.

How to use

Check an article

1. Keep Zotero running. Standard mode uses Zotero's built-in Local API. In enhanced mode, the disabled Paper Library Checker (<version>) item in Zotero's Tools menu confirms that the add-on has loaded.
2. Open an article-detail page covered by Supported sites and status.
3. Wait for a status badge near the page title or in the lower-right corner.

The extension extracts page metadata and asks the selected backend to match it. Standard mode normally uses a minimal IndexedDB snapshot of the personal library and accessible group libraries. A ready snapshot gives complete exact identifier/title results without one Zotero HTTP search per candidate. While no index exists, detail checks and small batches use the Direct Local API compatibility path; its positive matches are exact, but quicksearch misses are incomplete. Enhanced mode uses the add-on's local in-memory index. The floating ↻ button re-checks an article, manually starts a batch check on supported list pages, and can be dragged.

Options includes a Standard mode index section with status, item/library counts, last successful update, progress, Refresh index, Clear index, Clear and rebuild, and Cancel build. The extension refreshes snapshots older than 30 minutes in the background. The previous complete generation remains queryable during refresh and is replaced only after the new full snapshot succeeds.

Toolbar popup

Select Paper Library Checker in the browser toolbar to see Zotero connection state, the actual active mode, matching/index capability, current-page state, and any automatic fallback reason. When repair is needed, Fix connection opens the relevant Options entry point. The legacy connected and indexReady health fields remain for compatibility; for Direct mode, indexReady means that the legacy health check is usable, not that a complete persistent index exists. New callers should use capabilities.engine, indexState, and the explicit completeness capabilities. Check this page uses the same manual-check entry point as ↻.

Unsupported page means the extension has no content script on the active tab, such as a browser-internal page or a website outside the manifest site list. It is not a new site-support claim. Browser UI follows the browser display language; the Zotero Tools menu follows the Zotero/Gecko locale. English and Simplified Chinese are included.

Status legend

Page badge	Meaning
Library: checking	A check is in progress.
Library: saved	A matching item was found in the local library.
Library: possible match	Enhanced mode found a fuzzy match that requires manual confirmation. Standard mode does not produce this state.
Library: not saved	No match was found using the metadata supplied by the current page.
Library: no match; result may be incomplete	Direct fallback found no candidate; build or refresh the standard index for a complete result.
Library: saved in previous index	A match exists in the previous snapshot while a refresh is needed or running.
Library: index needs update	The previous snapshot did not match; the extension will not claim an absolute absence until refresh completes.
Library: unrecognized	No supported metadata was recognized.
Library: choose item	translation-server returned multiple candidates.
Library: offline	The extension could not connect using the selected mode.
Library: indexing	The enhanced-mode local index is not ready yet.

Badge and page-glow colors use red for saved/matched, orange for possible matches, blue for a complete not-saved result, yellow for checking/unrecognized/choice/incomplete results, and purple for offline/indexing/error. Library: not saved is shown only for a complete ready-Indexed or XPI result using the metadata supplied by the page; it is not absolute proof about the entire Zotero library. Direct and stale misses use the qualified incomplete message.

Re-check after saving and list checks

Enhanced mode listens for Zotero item additions, modifications, deletions, and trash events and updates its add-on index in real time. Standard mode uses full snapshot refreshes because Zotero 9.0.6 does not provide a reliable complete since-based change guarantee. A successful standard refresh notifies previously checked pages to re-check. Use Refresh index after a change when an immediate result matters.

Scopus /pages/publications/<id> document-detail pages are checked from their visible DOI, title, displayed author names, year, and source metadata. CNKI KNS8 and Scopus publications search-result pages are checked automatically and display inline positive matches; Scopus Table and List views are supported. Saved and possibly saved rows are highlighted by default. Auto-check reference lists remains off by default and controls reference/citation lists: when enabled, supported lists are checked as they load and scroll; otherwise click ↻ to start a supported list check manually. A page processes at most 80 candidates. Scopus detail and search checks and CNKI search checks remain experimental, ScienceDirect is best effort, and MDPI References are not supported.

Web of Science /wos/<database>/full-record/<id> pages use the visible Full Record title, DOI, full author names, publication date, document type, and source. /wos/<database>/summary/<search-id>/<sort>/<page> result pages are checked automatically from their stable record markers; when available, each result's OpenURL supplies its DOI and unabbreviated first-author name. Only positive matches receive inline badges and row highlighting. Both Web of Science integrations remain experimental and were verified against the live 2026 Angular DOM.

Visual cues and pairing-token actions

Options includes a Visual cues section. Search-result row highlighting is on by default, with separate colors for saved and possibly saved results. The page edge is off by default; when enabled, its result colors, 1–16 px width, and glow, solid, dashed, dotted, or double style can be customized. Saving or restoring these visual settings updates open supported pages immediately. Glow animation respects prefers-reduced-motion. These controls change presentation only and do not change matching results.

• Copy pairing token copies the current token.
• Reset pairing token generates and automatically copies a new token; the old token stops working.
• Revoke pairing token immediately revokes the current token.

After Reset pairing token, paste the new token in extension Options, click Save, and then Test connection. After Revoke pairing token, the extension cannot connect until a new token is generated and saved.

Common problems

Problem	What to check
No badge appears	Confirm that Zotero is running, the add-on is loaded, and the extension is enabled. Confirm that the domain appears in the support table and that the page exposes usable citation, DC, COinS, JSON-LD, or CNKI metadata. Refresh or click ↻.
Library: offline	Start Zotero and select Test connection. For standard mode, enable Zotero Local API. For enhanced mode, keep the default endpoint and copy the token again if necessary.
Library: indexing	Wait for the local index and click ↻; restart Zotero if the status persists.
Library: possible match	This is a fuzzy title match, not a confirmed saved item. Compare the title, year, and authors in Zotero.
Library: unrecognized	The page did not provide usable supported metadata. PDF pages are especially likely to lack enough metadata.
Reference links are not colored	Enable Auto-check reference lists or click ↻, and confirm that the page has a supported list adapter.
A manually loaded extension disappears after restart	The unpacked extension directory must remain in its original location. Use Load unpacked again and pair again if it moved.

Update or uninstall

Chrome Web Store installations update through the Chrome Web Store channel; users do not need to download a ZIP to replace the store version. Downloading a GitHub browser ZIP and clicking Reload in chrome://extensions or edge://extensions applies only to manual/unpacked installations. Zotero XPI updates continue through the current GitHub/Zotero update flow. Removing the browser extension clears its local extension identity and storage, so a later reinstallation may require pasting the pairing token again and testing the connection.

To uninstall cleanly:

1. In Zotero, choose Tools > Paper Library Checker: Revoke pairing token.
2. Remove the browser extension from chrome://extensions or edge://extensions.
3. Remove the Zotero add-on from Tools > Plugins.
4. If it was installed manually, delete the unpacked browser-extension directory after the browser no longer lists it.

Users upgrading from 0.4.x should read the dual-mode migration guide. Users upgrading from a 0.2 development build should also follow the 0.3 migration.

How it differs from Zotero Connector

Zotero Connector saves items into Zotero. Paper Library Checker does not replace it and does not save items. It checks whether page metadata appears to match an item already in the local library, then displays a page badge or list marker so you can decide whether to save.

FAQ

Does Paper Library Checker work with CNKI?

Yes. CNKI Chinese article details are supported and tested. KNS8 search results have an explicit adapter; search results, CNKI English details, and CNKI reference/list checks remain experimental.

Does it upload my Zotero library?

No. Standard mode reads only matching candidates through Zotero's built-in Local API and processes raw records inside the extension service worker. Enhanced mode uses the add-on's authenticated local in-memory index and returns minimized results. The project does not upload Zotero library data or use it for telemetry. Optional translation-server integration sends only the current public page URL to a separately installed local service.

Is this a plagiarism checker?

No. Paper Library Checker checks whether a bibliographic item already exists in Zotero. It does not inspect paper full text, calculate text similarity, or detect plagiarism; “Zotero duplicate checking” here means duplicate saved-item detection only.

Local data flow and privacy

Scholarly page DOM -> content script -> extension service worker
    standard -> Zotero built-in Local API -> service-worker verification
    enhanced -> authenticated add-on API -> add-on in-memory index
Both return only status / match type / confidence

Candidate metadata may include title, public identifiers, date, limited creator values, and the current article URL. Matching stays on loopback, and responses do not expose Zotero item IDs, keys, stored URLs, attachments, notes, collections, or unrelated library metadata.

Standard mode requires no token. Raw Zotero Local API item JSON stays in service-worker memory and is not returned to the webpage, stored, logged, uploaded, or used for telemetry. Enhanced-mode pairing secrets remain in chrome.storage.local, not sync storage; its add-on requests use versioned HMAC-SHA256 authentication and minimized responses. Legacy bearer-token and token-in-JSON requests fail closed. Optional translation-server integration sends only the current public page URL to a separately installed service at 127.0.0.1:1969. Page badges remain observable by the visited page.

See PRIVACY.md, the threat model, and SECURITY.md. Do not report vulnerabilities in a public issue.

Advanced configuration

Options intended for advanced users

• endpoint: keep the default http://127.0.0.1:23119/zotero-checker.
• connectionMode=auto: prefer a healthy compatible enhanced backend, then fall back to standard mode and report the reason.
• connectionMode=standard: use only Zotero's built-in Local API; do not probe the add-on.
• connectionMode=enhanced: use only the authenticated add-on; do not fall back.
• translationServerMode=off: never use translation-server.
• translationServerMode=auto: try translation-server only when needed on priority academic domains; fall back to the local extractor if it fails.
• translationServerMode=always: try translation-server first.
• enablePageGlow: visual result cue only; default false.
• autoCheckReferenceLists: automatic supported-site batch checks; default false. Manual ↻ checks remain available.
• broadPageDetection: article-detail detection only where the manifest already injects the extension; default false. It does not expand host permissions.
• developerMode: display a bounded, service-worker-memory log in Options with backend phases, request timings, cache use, batch counts, and stable errors; default false. Refresh or clear it from the same panel. Pairing tokens and raw Zotero item records are never included, and restarting the service worker clears the log.

ScienceDirect and MDPI pages with citation_doi normally use the generic extractor in auto mode. MDPI References are not scanned. See matching behavior, architecture, and the local protocol for implementation details and limits.

Release verification

Release verification and SHA-256 checksums

Version 0.5.0 is the current GitHub alpha release. Its accepted qualification covers the canonical automated suite and the recorded Chrome, Zotero 9.0.6, Standard Indexed, Direct fallback, Enhanced/XPI, lifecycle, and CNKI runtime evidence. The public Chrome Web Store remains at 0.4.1 until its separate store update is published. The historical broader Edge qualification target remains Microsoft Edge 151.0.4129.78.

Download the public artifacts and SHA256SUMS.txt from the v0.5.0 GitHub Release. The repository-root updates.json publishes the matching 0.5.0 Zotero update URL and XPI hash. See the 0.5.0 qualification for the accepted candidate evidence.

Development

Requires Node.js 20.19 or newer.

npm ci
npm test
npm run check
npm run build
npm run inspect:artifacts
npm run verify:release

The PowerShell compatibility entry remains available as .\scripts\package-zotero-plugin.ps1. Synthetic fixture policy and manual site checks are documented in docs/test-matrix.md. Architecture and protocol details are in docs/architecture.md and docs/protocol.md.

Contributing

Read CONTRIBUTING.md. Current priorities are security review, synthetic adapter regressions, and reproducible release validation. User-visible changes are recorded in CHANGELOG.md.

License

Copyright 2026 he-chun. Licensed under the Apache License, Version 2.0. See LICENSE, the public license decision record, and THIRD_PARTY_NOTICES.md.