Scope, identity, and sharing
This policy explains how Nintek (“we”, “us”) handles information inTabloom. Tabloom is a private, single-owner notebook workspace that can grant selected guest access by email. It has no public self-service sign-up.
Microsoft identity provides the tenant and object identifiers, name, and email needed for access. Tabloom records first-seen and last-seen activity and uses email for sharing grants. Offline copies are identity-scoped but are not end-to-end encrypted.
Tabloom does not use third-party advertising or cross-app tracking and does not request precise location, contacts, or advertising identifiers.
Information Tabloom processes
- Identity and sharing. Microsoft tenant and object identifiers, profile name, lower-cased sign-in email, first-seen and last-seen timestamps, and email-based notebook grants.
- Workspace content. Notebooks, pages, rich-text blocks, links, tags, shares, archive and Trash state, versions, generated content, and settings.
- Uploads and scans. Images and files you add, Inkbox input, OCR text, image-processing output, and references needed to return the content.
- Operations. Request timing, errors, version information, security activity, AI-call status, quota and estimated-cost records, backup manifests, and recovery checks.
Anthropic, Voyage AI, and Azure image processing
Tabloom uses different processors for distinct tasks:
- Anthropic processes selected text, retrieved note context, or a chosen scan for summarizing, rewriting, Ask Your Notes, suggestions, and related generation.
- Voyage AI creates semantic-search embeddings from page text and search queries. The resulting vectors are stored with the private workspace.
- Azure Computer Vision Read API can extract text from an uploaded image when OCR is configured and you invoke that workflow.
- Azure OpenAI Service can use the configuredgpt-image-1 deployment to clean a selected Inkbox handwriting scan and place the result into a PDF. This narrow feature is disabled when its Azure configuration is absent.
Only the content needed for the selected operation is sent. Generated or extracted output can be wrong; compare it with the original before relying on it. Provider-side processing and retention follow the provider’s terms.
Server storage and explicit offline reading
Tabloom stores its owner-and-guest workspace in SQLite on a persistent Microsoft Azure App Service volume. Upload bytes live in an account-scoped uploads directory referenced by the database; they are not authoritative live media in Azure Blob Storage. HTTPS protects service traffic in transit.
Tabloom is installable but is not local-first. When you explicitly download a notebook, or when a bounded recent-page setting keeps an eligible page, sanitized read-only copies are stored in identity-scoped IndexedDB. The service worker caches the public shell and static assets, not authenticated API responses.
Offline editing, creation, deletion, sharing, scanning, AI, Search, Inbox, Trash, live History, and server-backed settings remain unavailable. After reconnecting, the workspace stays read only until a complete authoritative refresh succeeds. Anyone with access to the browser profile may be able to read offline copies, and the browser may evict them.
Export, Archive, Trash, history, and account deletion
Tabloom provides page, notebook, and workspace export routes. Workspace ZIP is the default and contains Markdown pages, indexes, links when selected, and referenced or Media-tab attachments by default. HTML is also available; the PDF choice produces printable HTML for browser print-to-PDF. Exports do not include the underlying Microsoft account, every provider record, or operator backup history.
Archive hides content from the active workspace but does not delete it. Trash is a soft-delete state: a daily sweep permanently removes items older than 30 days, and owners can permanently delete supported content sooner. Page history retains the most recent 50 versions for each page.
There is currently no self-service Tabloom account export-and-delete flow and no account-deletion API route. Emailing a request starts a manual review; it is not a promise of immediate automated deletion and does not delete the Microsoft account or independent provider records.
| Data or control | Where it lives | Export or backup | Deletion and retention |
|---|---|---|---|
| Identity, email, sharing, and activity | Microsoft identity supplies tenant and object identifiers, name, and email. Tabloom stores first-seen and last-seen activity and sharing grants by email in its server database. | Workspace exports include content, not the underlying Microsoft account or every provider security record. | There is no self-service account-deletion route. A manual request can be reviewed for Tabloom-held records; it does not delete the Microsoft account. |
| Notebooks, pages, archive, Trash, versions, and generated content | A server-side SQLite workspace database hosted on Microsoft Azure. This is a shared owner-and-guest workspace, not a separate database for every identity. | Page, notebook, and workspace exports support ZIP, HTML, and browser print-to-PDF. ZIP includes Markdown and, by default, referenced and Media-tab attachments. | Archive is not deletion. Trash is permanently swept after 30 days; owners can permanently delete sooner. Page history keeps the most recent 50 versions per page. |
| Uploads and Inkbox media | Account-scoped files on the Azure App Service persistent uploads volume, referenced by the SQLite database. | Exports include only the media supported by the selected export format. Keep independent originals for files that must survive outside Tabloom. | Permanent content deletion removes associated active records and files as implemented. Recovery bundles can retain older copies until their retention expires. |
| Explicit offline reading copies | Sanitized, identity-scoped IndexedDB in the browser profile. Remote images are stripped. Copies are not end-to-end encrypted. | The copy is a bounded read-only notebook or recent-page download, not a full editable workspace export. | Clear the downloaded notebook or site data, change identity, or complete sign-out. Sign-out is blocked if Tabloom cannot clear the private offline stores. |
| AI, embedding, and OCR inputs and outputs | Scoped content can be sent to Anthropic, Voyage AI, or configured Azure image-processing services for the selected operation. Resulting text, vectors, and usage records can be stored with the workspace. | Saved generated content can appear in content exports; provider-side operational records are controlled by the provider. | Delete the saved content or request review of account-associated records. Provider retention follows each provider’s terms and legal obligations. |
| Recovery and support records | Verified local recovery bundles include SQLite and uploads and are retained for 14 days. Optional off-host Azure export is a separate, off-by-default recovery control. Support mail exists only if you contact Nintek. | Operator recovery bundles are not a user account export. You control any support message you send. | Deleted data can remain in local backups until expiry. Nintek reviews support-mail requests subject to response, security, and legal needs. |
Scroll horizontally to read all four columns →
AI quotas, backups, and operations
AI and image operations are admitted per Microsoft principal. The current limits are 30 requests in a rolling hour and a $5 estimated monthly processing budget. A quota failure stops the requested operation rather than silently switching identity or charging another user’s allowance.
Tabloom creates verified daily recovery bundles from SQLite and the uploads tree and retains local bundles for 14 days. A managed-identity off-host exporter exists but is off by default and remains an operator recovery control, not a user export or a promise that every deployment has an activated off-host copy.
Children, changes, and contact
Tabloom is intended for a general adult audience and is not directed to children. We do not knowingly create Tabloom access for children.
We may update this policy when the service, processors, quotas, offline scope, retention, or deletion controls change. The dates at the top identify this version.
Privacy, access, and manual deletion requests can be sent toprivacy@nintek.com. Product help is available on Tabloom Support. Use of Tabloom is subject to the Nintek Terms of Use.
Evidence note. Reviewed against immutable Tabloom release-visual authority acfa8a9 and its matching live deployment on August 23, 2026. Account deletion remains a manual process because no self-service deletion endpoint exists.