The short version
Seba is anonymous by default. You can read the public library, ask Sebastian, and use the dream room without giving us your name, email, phone number, or payment details. We do not render your name, photograph, or email anywhere in the public interface. We do not sell data. We do not run third-party trackers or advertising pixels. The site works without sign-in.
Without sign-in, Seba can still keep browser-level continuity: conversations and saved items are tied to a randomly generated identifier stored in your browser. Sebastian’s private rolling memory starts when you sign in. Google sign-in links your records to your account so they follow you across devices. Optional donations, newsletter signup, feedback, or support emails collect only the information needed to handle that specific request.
What we store on the server
These categories may live on the Seba server, keyed either by the browser identifier on this device or by your account if you sign in.
Chat threads. Your conversations with Sebastian — the messages, the order they were sent, and timestamps. So you can come back on the same browser and pick up where you left off.
Saved readings and notes. Dream interpretations, tarot / I Ching / astrology readings you saved, and any Sebastian responses you bookmarked. Without sign-in, these are primarily device-local. When you sign in, they can sync to your profile.
Profile memory for signed-in readers. A short prose summary Sebastian writes about you over time, so the next conversation is not a cold start. It draws on your signed-in conversations across the site — chat with Sebastian and the experience flows (dream, tarot, astrology, and I Ching readings) can all contribute to it. It is short by design, lives in your profile record only, and is not created for signed-out readers.
Anonymous dream-theme counts. We keep anonymous daily counts of common dream themes (like teeth or water) to understand what themes people bring — never the words of a dream, and never tied to a person or account.
The memory records themselves do not store your IP address, and they are not visible to other users. Separate service logs (described below under “What we collect”) may use a short-lived IP-based key for rate limiting and reliability. We do not sell or share any of this.
What we collect
Analytics. We run a self-hosted instance of Umami to understand how pages perform. It records page URL, referrer, screen size, browser, and country — nothing that identifies you personally. It does not set cookies. It does not log your IP address in identifiable form.
Campaign attribution. When a campaign link brings you to Seba, your browser may keep a bounded first-touch record for up to 30 days so we can understand whether that visit later led to a reading, saved item, membership, or donation. The record contains only allowlisted campaign labels and the landing-page path. It never contains your dream, question, reading, full URL, query string, or referrer string, and it stays in your browser’s local storage.
Sign-in (optional). If you sign in with Google, we receive your Google account identifier, email, display name, and profile-photo URL from Google’s sign-in service. We use the identifier to link your records to your account so they follow you across devices. We do not render your name, photograph, or email anywhere in the public interface. We do not email you directly without your opt-in through the newsletter form.
AI processing. When you ask Sebastian a question or use an experience, the text you write and the context needed to answer it — including selected library passages — are processed by third-party AI providers to generate the response, subject to those providers’ API data-use terms. Depending on the feature, Seba uses or has used DeepSeek, Anthropic (Claude), OpenAI, Voyage AI, Cloudflare Workers AI, and DeepInfra.
Voice dictation (when offered). If you tap the microphone in Sebastian, the audio is processed by Cloudflare Workers AI using a speech-to-text model. Seba does not save the recording. The resulting transcript is placed into the text box for you to review and edit; it is not sent to Sebastian or saved as part of a conversation unless you press Send.
Spoken answers (when offered). If you tap Listen beneath a completed answer, the visible answer text is sent to DeepInfra’s Kokoro text-to-speech service. Free visitors may receive a limited listening sample; membership raises the listening allowance. Seba does not send your name, email, member credential, source packet, or conversation history with the text. Generated audio is not stored on Seba’s servers; your browser may keep a private local copy so replaying the same answer does not generate it again.
Seba measures the free listening sample with a one-way digest of the anonymous browser identifier for about one year. A short-lived one-way digest of the network address limits repeated resets; the raw address is not stored in the speech counter. These counters contain character totals and dates, not answer text or audio. Anonymous analytics may record that a sample started or finished, or that a membership link was opened, without recording what was read.
Seba does not attach your account name, email, or other profile identity to these requests. Anything you include in the text itself, however, is sent as part of the request. No account, name, or email is required to use Seba. On Seba’s side, activity is associated with an anonymous identifier stored in your browser. Dream text is not stored on our servers unless you choose to save a reading. Chat conversations are kept under the anonymous identifier so you can return to them on the same browser; you can delete them or ask us to delete them. If you sign in, your saved items and conversations are linked to your profile instead.
We are building a local-processing option so this work can run on infrastructure we operate rather than third-party AI services. Check back here for updates. Section updated August 2026.
Newsletter (optional). If you subscribe to our newsletter, we store your email address with our email provider so we can send you occasional dispatches. You can unsubscribe at any time from any email we send.
Donations (optional). Donations are handled by
Stripe. Seba does not see or store your card number. To recognize
supporters and suppress donation nudges, the site may send Stripe
the anonymous browser identifier stored under
seba.anon_uuid, then store a supporter flag keyed to
that identifier and, when available, a short-lived IP-based key.
Seba also stores a Stripe Checkout Session reference for
reconciliation. These supporter records are kept for about one year.
Your card statement should show SEBA.HEALTH or similar,
depending on your bank.
Estate membership (optional). Membership checkout, recurring billing, and the manage-or-cancel portal are hosted by Stripe. Seba stores an opaque member key, your anonymous Seba profile identifier, the Stripe subscription and customer references needed to recognize access and open that portal, and the membership start date. Seba does not store your card number or copy Stripe’s billing identity into the estate ledger. Stripe retains the billing records required to operate the subscription.
Dream-reading credits (optional, when offered). One-off credit checkout is hosted by Stripe. Seba sends an opaque browser identifier, then stores the verified Checkout Session marker and credit balance against that browser identifier and, when available, a short-lived IP-based key. Credits expire after about ninety days. Seba does not store your card number or copy Stripe’s billing identity into the credit record.
Feedback / Report a problem. If you use the “Report a problem” form on /profile or the site-wide feedback banner, we store the message you wrote, the optional contact email if you provided one, the page URL you were on, your browser’s user agent, and your profile identifier if one is present in your browser. We use this to debug the problem and to reply to you if you left an email. Reports are stored in our support queue and may be forwarded to our team’s inbox; we do not share them.
Service logs. When you ask Sebastian or use any of the public AI endpoints (chat, dream, tarot, I Ching, astrology, matchmaker), the backend may store a short request label or, on the public Oracle / search service, query text; a timestamp; an anonymous browser identifier when available; response-status or compact error information; and duration / cost metadata. The dream room does not put the words of your dream into these service logs. Separate rate-limiting and reliability controls may temporarily use a short-lived IP-based key. We use these records for rate limiting, abuse prevention, debugging, and keeping the service reliable. Service logs are kept separate from your profile memory; they are not used to enrich Sebastian’s memory of you, and your profile memory does not contain an IP-based key.
On-device storage. Your browser stores a session
token (if you signed in), your reading palette preference, and an
anonymous identifier under the key seba.anon_uuid.
That identifier is the only thing that lets your browser retrieve
browser-linked chat threads and saved items from the server.
For your current dream reading, the browser may also keep a
device-local recovery copy for up to four hours so a refresh or
phone suspension does not erase the intake or a reading that finished
while the phone was asleep. It is removed when you start over, sign
out, clear browser storage, or the four-hour window expires.
Clearing your browser storage removes the identifier;
the server-side records remain but become unreachable from that
browser. To remove them, see “How to delete your data” below.
How long we keep data
Profile memory, saved readings, notes, and browser-linked threads remain until you delete them or ask us to delete them, because their purpose is continuity. Service logs are shorter lived: usage events and public Oracle query logs are purged after about 90 days; expired sessions and magic sign-in links are purged after about 30 days; forwarded feedback reports are purged after about 12 months. Donation/supporter records are kept for about one year, as described above. Dream-reading credit balances and their redemption markers expire after about ninety days. Aggregate daily-cost rows are not personal records.
Practitioner interview submissions are kept for editorial review and removed when a correction/removal request requires it; public directory pages are handled through the directory correction and removal process.
What we don’t collect
No advertising cookies. No fingerprinting. No session recording. No third-party analytics (Google Analytics, Facebook Pixel, etc.). No sale of data to any party.
How to delete your data
Anonymous visitors. Your records on the server
are keyed only by an opaque identifier — we have no way to
look you up by name or email. To request deletion, email
cody@seba.health and include
your identifier. You can find it in your browser’s devtools
under localStorage, the key seba.anon_uuid. We will
delete every record under that identifier.
Signed-in accounts. Email cody@seba.health from the address you signed up with and we will delete the account and all associated records.
Payment, supporter, and reading-credit records. If you donated or purchased a reading credit and want the associated records removed, include the browser identifier and, if you have it, the Stripe receipt or Checkout Session reference. Records linked to the browser identifier can be removed. IP-keyed marks are only removable when the exact key is identifiable; otherwise they expire automatically.
Newsletter. Use the unsubscribe link in any email we send.
Clearing your browser’s site storage for seba.health removes the identifier from this device but does not delete the server-side records on its own — without the identifier we can’t find them to delete them. If you want them gone, send the identifier first.
Contact
Seba is operated by 702 Alliance LLC, doing business as Seba Health. Questions about privacy go to cody@seba.health. We respond to privacy and deletion requests within 30 days.