1. What data the extension processes
When you click "抓取到百晓" (or "Capture to Baixiao") on an article page, the extension reads the following from the active tab:
- The article's bibliographic metadata (title, authors, publish year, journal name, volume / issue / pages, DOI, abstract) — extracted from page meta tags and DOM elements.
- The article's source URL.
- The article's PDF bytes, fetched from the publisher using your existing browser session (including any institutional-access cookies).
The extension does not process or transmit:
- Your browsing history outside of articles you explicitly capture.
- Cookies from publisher sites — those remain in your browser and are only used to authenticate the PDF fetch.
- Any data while the extension is idle (i.e., when you are not actively clicking the capture button).
2. Where the data goes
Captured articles (metadata + PDF bytes) are uploaded to the Baixiao API server at https://chat.know-pa.cn (or any alternative server URL you configure on the extension's settings page) and stored in your Baixiao knowledge base. The upload is associated with your Baixiao user account via a bearer token.
Your bearer token is stored in your browser's local extension storage (chrome.storage.local). It is transmitted only to the Baixiao API server, never to any third party.
3. Third-party services
For DOI-keyed articles, the Baixiao backend may call the following public APIs to enrich missing bibliographic fields:
- Crossref (
api.crossref.org) - OpenAlex (
api.openalex.org) - Unpaywall (
api.unpaywall.org)
Only the article's DOI is sent to these services. No personally identifiable information is included in those calls.
4. Institutional-access cookies
To support paywalled articles via your institution's subscription, the extension's background service worker uses your browser's existing cookies for the publisher's domain (e.g., wiley.com, sagepub.com) when fetching the PDF. This means:
- Articles you can already access in your browser, the extension can capture for you.
- Articles you do not have institutional access to remain inaccessible.
- Publisher cookies are read only at the moment of fetching the PDF you explicitly capture. They are not exfiltrated, logged, or sent anywhere outside the publisher's own domain.
For the CNKI database specifically, the extension uses a per-request Referer header override (via chrome.declarativeNetRequest) to bypass CNKI's anti-leech check on PDF downloads. The override is applied only to the specific PDF URL for the in-flight request and is removed immediately after.
For the Wanfang database specifically, the extension intercepts Wanfang's own download flow (via chrome.downloads.onCreated) to capture the PDF bytes for upload to your knowledge base.
5. Permissions used
Permission Why it's needed
activeTab Read the article page you're viewing so the extension can detect metadata and the PDF URL.
scripting Inject the metadata-detection script into the article page.
storage Persist your Baixiao bearer token and last-used knowledge-base ID on-device.
notifications Notify you when async PDF processing (OCR) completes after upload.
downloads Intercept Wanfang database downloads to capture PDF bytes for upload (otherwise Wanfang PDFs save to disk and the extension cannot see them).
declarativeNetRequestWithHostAccess Inject the Referer header for CNKI PDF fetches to pass CNKI's anti-leech check. Scoped to one URL per in-flight request.
host_permissions: https://*/* Fetch publisher PDFs cross-origin using your session cookies — required to support any institutional-access publisher.
6. Your data, your control
- Delete a captured article: from your Baixiao knowledge base — this removes both the PDF and the metadata server-side.
- Revoke the extension's access: regenerate your bearer token at account settings, or simply uninstall the extension.
- Export your data: available through your Baixiao account's standard export tools.
7. Children and sensitive categories
The extension is not directed to children under 13. It does not knowingly process special-category personal data.
8. Changes to this policy
We may update this policy as the extension evolves. Material changes will be reflected here with an updated "Last updated" date and announced in the extension's update notes.