Sign-in and personal data
Applies to leafhour.com · Updated October 22, 2026
Leafhour uses your sign-in identity to confirm who a reading room belongs to, and stores the books you choose to upload together with your reading records. Public samples can be read anonymously. This notice explains how this online instance handles data; a self-deployed instance is the responsibility of its operator.
What sign-in uses
Email sign-in uses the address you enter. Supabase Auth verifies the password or code, and Resend delivers verification email. Sign-up and password recovery verify your email first, then set a password. When you choose Google or Apple sign-in, that provider verifies your identity; Supabase receives the account identifier, email address and the basic profile the provider shares, such as name and avatar. Leafhour uses this information to recognize your account and keep you signed in. It does not request access to email contents, contacts, cloud drives or calendars.
When Apple provides a hidden address, the account is stored under that identity. An official authorization flow starts only after the sign-in page shows that the method is available. You can revoke Leafhour’s access in your Google or Apple account settings.
Provider access and refresh tokens are kept on the server. The browser receives cookies used only for sessions on this site; a password sign-in session lasts at most seven days, and the temporary verification cookie for third-party sign-in lasts at most ten minutes. Signing out revokes the session on this site; it does not delete reading room data.
Your books and reading records
When you choose to upload, the file and its extracted source text are stored in your private reading room. Comments, questions, reading position, reading time, explicitly marked completion and saved translations are managed per account. Online storage is provided by Supabase and the website is hosted on Vercel. Anonymous public books and private uploads are handled separately; private files and records are never made public by uploading them.
Once reading starts, the page counts only time when it is visible and in the foreground. Background time pauses, and long gaps while the browser is suspended are not counted; reading time does not measure understanding. The calendar is shown in your local time zone, and time across midnight is recorded separately. The browser caches reading settings and some reading state so you can continue; while offline it holds pending reading sessions, chapter marks and unconfirmed submissions for this account, which can be retried once you are online again. Text you have not saved stays as a local draft. Selecting text, switching translations and adjusting typography happen locally; paragraph context is sent to the configured model service only when you explicitly request an AI translation or explanation. Per-paragraph translation currently uses DeepSeek and sends only the selected paragraph and the adjacent passages needed to understand it. The reading conversation may consult this account’s reading room, public books and web material on request; these tools are read-only, do not read private notes, do not write reading progress, and never buy books or add recommendations to your shelf. Sign-in details are not used as translation context.
If you authorize your own agent, it can only read the books and data types you grant; write access is chosen separately. Grants created by older clients still work; to revoke one, use that client or contact the address below. Revocation stops future access but cannot automatically erase data the other side has already copied.
Reading points and payments
If you choose a monthly or yearly subscription, Creem hosts checkout, processes each payment and provides the authenticated customer portal for subscription management. Leafhour associates your account ID with local order, checkout, transaction, subscription, customer and refund identifiers, product, currency, amounts, payment and subscription status, billing periods and an immutable credit ledger. Creem receives the contact and payment information needed for checkout. Leafhour does not receive or store full payment-card details. Private books, questions, answers and notes are not sent to Creem.
Provider usage records keep operation and attempt IDs, feature, model, reported token usage and available cost estimates separately from reading content. Missing supplier usage remains unknown. Operator billing reports use necessary financial and usage facts, not private book text or notes. Financial records may need to be retained for payment reconciliation, disputes and legal obligations after reading data or an account is deleted; deletion requests are assessed separately for these records. Free reading and export do not depend on buying points.
Pricing · Terms · Refund policy
Storage, export and deletion
Reading room data is kept so you can resume reading and export it. A reading conversation can be exported from its window, with AI answers and your own words marked separately. Library → Settings → Data and migration offers a complete reading-library ZIP backup and restore, selected personal records in Markdown/JSON, and a sharing preview, without a subscription. It preserves synced source text, translations, comments, conversations, citations, positions and time records. Original files are optional. Credentials, Agent grants and unfinished model jobs are excluded; listening audio and listening histories use the separate audio export. Unsynced reading records and unsaved drafts stay in the original browser; finish syncing or saving them first. To delete an account or private data, contact the address below; we verify account ownership before handling the data. Logs and backups that the hosting services produce for operation, security and troubleshooting are handled under those services’ own retention policies.
Sign-in data is used for authentication, sessions, security and the reading features you request. It is not used for advertising and is never sold. Supabase, Resend, Vercel, identity providers and the model services you actively call process the data needed for their features and are governed by their own privacy policies.