Shiye (Page Memory) privacy policy
Shiye helps you keep pages locally so you can find them later. We minimize permissions and prioritize local storage: installation does not read every site you visit.
1. What we collect
The extension may keep the followingStored only in your local browserinformation:
- URL, title, site name and author when extractable
- Extracted page text, summaries, topics and visit times
- After you opt into tab memory: URLs, titles, first/last opening times, up to 32 opening clues and pending/done status for current and future ordinary tabs. Reliable parent-tab URLs/titles may also be kept locally. Existing tabs have a collection time only when first enabled; earlier opening times are not inferred. An opening does not mean a page was read. Unreliable sources are shown as unknown.
- Your preferences: pause, excluded domains, reminders, custom tags, archive and star status
Local saving requires no account, name, email, payment information or precise location. Feedback contact information is separately optional, and feedback may be fully anonymous. Saved page text may contain personal or sensitive information: exclude email, financial, chat and other private sites. Forms and editable areas are removed before extraction, but not all sensitive content on a page can be identified automatically.
2. When pages are read
- Default mode: click-to-save: reads only after you click the extension icon or Save this page in the popupCurrent tabextractable content.
- Optional mode: all-site automatic saving: only after you explicitly enable all-site automatic saving in the popup or library and accept Chrome's website-access request does the extension save foreground HTTP(S) pages automatically.
- Optional tab-memory mode (off by default): after explicit opt-in, existing tabs access collects current and future HTTP(S) tabs, including background, inactive and unread tabs. It uses only metadata already provided by the browser: no loading, activation, background text injection or browser-history permission. Pages closed before installation are not recovered. Titles and URLs may also be sensitive; exclude sites as needed.
- Turning off tab memory stops new metadata but keeps text permissions. Turning off automatic text saving does not turn off tab memory. Global pause, exclusions and incognito restrictions apply to both modes. Enabling a mode does not resume paused saving.
- You can turn off all-site automatic saving and return to click-to-save at any time.
- Incognito, Chrome internal pages and excluded domains are not captured.
3. Where data is stored and how it is used
- Data is stored in your local Chrome extension storage/IndexedDB for display, search, review, export and import.
- Summaries and topics are generated by local rules. They are not uploaded to developer servers or sent to third-party cloud AI.
- The extension itselfWe do not uploadyour page text, history or settings to our servers; this version has no developer-operated cloud sync.
4. Permissions
activeTab/scripting: saves the current page when you click the extension.storage/unlimitedStorage: saves records and settings locally.tabs: recognizes tab state. After tab-memory opt-in, reads ordinary tab URLs/titles and parent relationships for later recall. Metadata mode does not read background page text.alarms/notifications: optional reminders to revisit yesterday's reading.optional_host_permissions(http/https): all-site access is requested only when you choose automatic text saving. When a feedback service is configured, feedback or a separate metrics choice requests only that service's exact site access, without extending text-capture scope.
5. Third-party code
Text extraction uses a locally bundled Mozilla Readability (Apache-2.0). It runs in your browser; this does not send content to Mozilla or to us.
6. Your controls
- Pause saving, exclude domains, delete records, clear data or uninstall at any time. Uninstalling removes local extension data. Single-record deletion keeps a local URL stop marker to prevent automatic recollection; an explicit manual save or import can restore it. Clearing removes text, indexes and opening clues and pauses both modes, while retaining minimal URL stop markers so open tabs cannot refill automatically. Explicitly recollecting current tabs can restore deleted links. Reopening never marks a record read automatically.
- Export/import JSON backups. Version 2 includes tab metadata, sources and status, and supports version 1 backups. Import skips existing records rather than overwriting newer content. Settings, mode switches and browser permissions do not migrate. Keep backup files safe.
- You can turn reminder notifications off.
Voluntary feedback and minimal product events (separate choices)
Tab memory, saved text and public downloads require no account, contact details or metrics consent. Help improve Shiye is off by default and separate from tab saving. API-disabled candidates do not send test requests to production. With a configured service, only after explicit consent do we send a random instance ID, extension version, platform category, the current consent event and the first subsequent tab-memory enablement event. Earlier behavior is not backfilled. URLs, titles, text, search terms, excluded domains, tags and total record counts are excluded. These are consenting samples, not all installs or unique people.
Withdrawal stops sends, aborts requests and clears pending events and the local instance ID. In-flight requests may already have arrived; existing server records are not guaranteed erased. The queue is capped at8 with bounded retry and a 3-second request budget. Failures do not block local features. Changing the service origin requires renewed consent and access.
Feedback can contain an anonymous category and description; email is optional for a human reply. An instance ID is attached only when you separately choose to link it, never inferred from metrics consent. Extension version and platform category are included. No attachments, automatic browser-data collection, email verification or automatic email sending. Do not include private browsing content. Failed submissions retain a local draft and request ID for manual retry; a real server receipt clears the raw draft, which can also be cleared manually. The receiving service and operator access must be protected. Production enablement and retention/deletion arrangements require separate review and disclosure. API-disabled local candidates do not send requests.
Clearing also creates local URL-fingerprint stop markers for ordinary tabs currently open, without retaining unsaved raw URLs. This prevents old tabs that were never saved while paused from refilling on resume. Only explicit recollection or manual saving removes the stop.
Website downloads and receiving service
Public downloads require no login or registration. Only separate consent on the download page sends a random click ID and version. Declining or service errors do not affect downloading. A click is not a completed download, install or person count. Website feedback can optionally include version/platform; extension feedback explains its included version/platform before submission. Optional email and instance linking are separate from metrics consent.
The receiving service hashes instance IDs with a secret key and stores feedback separately from counts. Only authorized operators can sign into the protected local management page; feedback contact information and metrics have separate access roles. CLI tools are supplemental. Status and internal notes do not email the sender automatically. Feedback and events are retained for 90 days by default; provide your feedback reference to request deletion or correction. No public feedback list or detail page is offered. IP addresses and device fingerprints are not used to identify people. Application logs do not record request bodies, email addresses or raw IPs.
The new service is currently a local candidate; website and extension APIs are off by default. Production proxying, backup retention/deletion, cleanup after recovery and operator access need separate review. This does not claim production enablement, off-machine backups or complete erasure from old copies. Existing Cloudflare website performance measurement is separately disclosed from voluntary product events.
7. Data sharing and sales
We do not sell your personal browsing data. Browser content stays local by default; the developer has no browsing-text database to sell or use for ad tracking.
8. Children's privacy
This extension is not designed specifically for children and does not knowingly collect children's personal information.
9. Policy updates
If privacy practices change materially, we will update this page and its effective date. Significant changes should also be noted in the store description or release notes.
Feedback and privacy contact
Use hello@shiye.app to contact us. When feedback reception is enabled, the trusted proxy temporarily parses source IPs for rate limiting. The application does not persist raw IPs; in-memory buckets retain only HMAC values made with a random process key, which expires on restart.
Website measurement
The product home, download, help and privacy pages on shiye.app use Cloudflare Web Analytics to understand page visits and performance. Your browser sends related measurement data to Cloudflare when visiting these pages. This script is excluded from the interactive demo workspace, library gateway and extension pages. It does not read saved extension text, search terms or private page URLs. Website measurement and the extension's local archive are separate data flows.
See About Cloudflare Web Analytics. The site remains usable if your browser blocks the analytics script.