Back to Omantia
EN

Privacy policy for Omantia

Updated on 25 October 2026.

Text approved for integration. Retention periods are pending verification before launch. This text does not replace a legal review.

1. Responsible organization and contact

Escuela Llenquén, Chile, is responsible for Omantia.

For privacy questions or requests concerning your data, write to escuelallenquen@gmail.com.

2. What information is processed and why

Omantia generates tarot readings using artificial intelligence for entertainment purposes. It does not replace medical, psychological, legal or financial advice.

To prepare a reading, we process the question you enter, the chosen topic, language and pronoun preference. You can request a reading without entering a question.

Avoid including names, contact details, medical information or other personal data about yourself or others.

Technical information necessary to deliver the service, prevent automated use and limit abusive requests is also processed, including the IP address in infrastructure controls.

When a reading offers the extended answer, the Worker seals that consultation in an encrypted and authenticated context your browser cannot read or verify. The page keeps it only in memory and sends it to the Worker only if you request the extended answer: it carries the question, the topic and structure already resolved, the cards with their orientations, the selections, the brief answer you were shown and the versions used, and it expires at most twenty minutes after it was issued. It is not stored in your browser, in technical logs, in metrics or in a database: if it was tampered with or has expired, the extended answer is not generated and you are asked to start another consultation.

After a reading, you may voluntarily submit a usefulness rating or a comment to help us improve. They are optional and independent. Do not include personal data or sensitive information: the text you submit may be retained even if it contains that information. In Tarot and horoscope comments, an initially unchecked checkbox lets you attach the brief or extended answer you are commenting on and your question, if any, for that submission only. Nothing is attached unless you check it. This content may be personal or sensitive, is recorded only in the corresponding feedback dataset and has the same intended three-month retention period; it is not sent to AI providers, logs or public errors.

When a mitigation notice is shown with a reading, you can report that the notice was wrong. That report always works without any text. If you tick the option to include them, it also carries the question you submitted and the answer you were shown. The option starts unticked, nothing is attached unless you tick it, and you can untick it and send the report alone. When a notice appears with an error instead of a reading, the same report is offered without that option: there is no answer to attach. The report records the notice shown, the mitigation context when the service is configured to include it, the language, the service version and your optional comment.

When you request a follow-up card, the service also processes the original question, complete reading and signed context, together with your optional new question. That context is held only in page memory and sent for server verification; a signature does not encrypt it. The additional card has no separate feedback form.

3. External services

We use Google Analytics to measure visits and navigation. It uses cookies with browser identifiers and processes technical device and connection information. We do not send questions, readings or feedback comments, and advertising features are disabled. See Google's privacy policy.

Omantia uses:

Turnstile security verification is subject to Cloudflare'sTurnstile Privacy Addendum.

The question and the data needed to generate the reading are transmitted to OpenAI through Cloudflare. These providers may use infrastructure and subprocessors located outside Chile.

Their terms and policies also govern the processing they carry out. We do not guarantee that processing takes place exclusively in Chile.

Voice dictation is optional and is provided by your browser. When you activate it explicitly on a compatible phone, the browser may send the audio to its own speech recognition provider to turn it into text; that processing is governed by the browser and its provider, not by Omantia. Omantia does not add its own transcription service, does not retain the audio and does not store the recognized text beyond the in-memory question field. The feature is not available on all phones: the icon appears only when the browser offers it and the screen is narrow, and the microphone starts only on an explicit click, with a visible indicator while listening.

The horoscope result cache is being prepared and is not yet activated. When enabled, Cloudflare KV will retain general horoscope text for a city, date, sign and language for 48 hours from writing, together with city identity, name, country, time zone, source and author references. This includes cities selected manually, but does not store a visitor's city preference on the server. It contains no visitor identifiers, IP addresses, coordinates, verification tokens or signed city descriptors. It does not permit storing Tarot questions or readings.

4. Storage in your browser

The horoscope city-search client is being prepared and is not yet activated on these pages. When enabled, an explicit city search sends the search text to Geoapify through our Cloudflare Worker, not directly from your browser. GeoNames supplies a local city catalog. No GPS permission or home address is requested.

An explicitly selected city may be saved underomantia_horoscope_city in local storage, as a signed city descriptor containing its identity, representative urban centre, time zone, display fields and data version. Search text and IP addresses are not saved in this preference. It remains until replaced, locally detectable corrupt or obsolete data is discarded, or you clear site data. If storage is blocked, the choice remains only in memory. This does not permit storing questions, readings or user identifiers.

The application stores your language and pronoun choices, using the keys omantia_locale and omantia_pronoun. You can remove them by clearing the site's data in your browser.

To vary the suggested questions, the application also keeps the identifiers of the three previous suggestions in the tab's session storage, using omantia_suggestion_previous. These reference the public catalog, not the text you enter or a user identifier.

The application does not automatically save your questions or readings in persistent browser storage. They are held temporarily in memory while you use the application. Your comment draft and its submission context are also held only in memory: closing the modal preserves the draft; starting a new consultation or reloading discards it.

This does not control features of your browser or device, such as screenshots.

5. Retention and metrics

Horoscope telemetry is being prepared and is not yet activated. When enabled, a dedicated Cloudflare Analytics Engine dataset, separate from Tarot and between environments, will retain operational cache/generation outcomes, sanitized error categories, duration and available provider token counts for three months. A valid new generation also records each planet and its author reference once; cached results do not. It records the consulted sign and, when resolved, the city identifier, name and country, with a separate global category that does not record the Greenwich reference as your location. This is the consulted city, not necessarily your residence or current location. It records no coordinates, IP address, city descriptor, verification token, request or user identifier, prompt or generated text. Missing or failed telemetry never prevents a valid horoscope. Dataset configuration and effective retention remain unverified.

The application does not maintain a persistent history of your questions or readings.

The metrics system is designed to record operational information and limited categories: language, topic, pronoun, selected cards, approximate country, general device characteristics, outcome, duration and service usage. It does not include the text of the question or interpretation automatically attached to an event.

Feedback is recorded separately in Cloudflare Analytics Engine. We accompany the rating or comment with the cards and orientations, topic, philosophers, concepts, narrative structure, language, model and versions used for that reading, and with the rated level when the reading offers a brief and a full answer: whether you rated the brief or the full reading. We do not include a consultation, user or session identifier, or an IP address, country or device in that event, and do not link it to the original consultation record.

Horoscope feedback is a separate, narrower exception. A horoscope reading may offer the same voluntary rating or comment tool, and we record it in its own dedicated Cloudflare Analytics Engine dataset with the consulted sign, the reading date you saw, the city identifier, name, country and resolved/global category authenticated for that reading, and the rated level when the horoscope offers a brief and a full version. Global feedback has no city location fields; older feedback may have no city. The consulted city is not necessarily your residence or current location. We never attach coordinates or the city descriptor. The question (if any) and reading text are attached only if you explicitly check the initially unchecked option for that comment; otherwise neither is sent. The event contains no verification token, signature, IP address, request or user identifier, language, model or version. Separate datasets are declared for production and preview; provisioning, deployed collection and effective three-month retention remain unverified. As with Tarot feedback, the comment is voluntary and may contain personal or sensitive information if you choose to write it; we never send it to an AI provider, a technical log or a public error, and the confirmation you receive means the submission was received, not that it is stored durably or can be retrieved or deleted individually.

The intended retention period for metrics and feedback is three months; effective retention remains subject to verification before enabling the service. Submissions are processed without a guarantee of storing or retrieving every event. Confirmation acknowledges receipt, not guaranteed retention.

This functionality is planned and is not active yet: the compact mitigation codes are recorded only when the service is configured to include them; if that configuration is not active, the event does not contain them. The intended disclosure for the planned stage is:

We record codes indicating which protections were included in the reading request and which notices were selected for the response. If you submit a rating, its context includes those same codes so we can analyze ratings by the mitigations applied.

These codes may reveal sensitive information relating to health, self-harm, legal or financial matters. They are not diagnoses and do not confirm your personal circumstances. They do not include your question or reading, and we do not add identifiers to link the two events individually. Encoding them does not make the information anonymous.

The planned retention period is three months.

The comment is a voluntary exception: it may contain personal or sensitive information if you choose to write or copy it. The same applies to a question and answer you deliberately attach to a report about a mitigation notice: they are sent only when you tick the option, are never attached automatically, and fall under the same planned retention period of three months, which remains subject to verification before the feature is enabled. We do not automatically attach your question or reading, or send the comment to AI services for analysis. Neither the comment nor an attached question or answer is recorded in technical logs or public errors, and an attachment is not part of the signed context. The absence of identifiers does not guarantee anonymity; the content and combination of metadata can be distinctive.

We also measure which catalog questions become visible and which are selected to improve suggestions. These metrics contain only the predefined question identifier, position, language and whether it appears among suggestions or in the full list. They do not include your entered text or user or session identifiers. Collection is indicative and may miss events.

This separation does not mean that no provider retains data. OpenAI may retain information in accordance with its terms and abuse-prevention controls. The technical option store: false does not guarantee zero retention.

Before publishing this policy, the effective retention periods for Cloudflare, OpenAI and metrics, as well as any special retention settings, must be confirmed and documented.

6. Your rights

You can contact Escuela Llenquén to request information about the processing of your data and exercise the rights of access, rectification, deletion or blocking that apply under the relevant legislation.

To handle a request, we may ask for proportionate information to verify your identity. Do not send identity documents or sensitive information in your first email.

If the application does not retain a consultation, it cannot retrieve it or delete a nonexistent record. Requests relating to information retained by providers will be handled according to the applicable responsibilities and procedures.

The interface does not allow editing or deleting submitted feedback. We do not retain a consultation or person identifier to locate each event and cannot guarantee individual retrieval or deletion in Analytics Engine. You may use the contact channel for privacy requests; this technical limitation does not replace an assessment of the rights and obligations that apply.

7. Security and changes

We apply controls intended to reduce unauthorized access and exposure of information. No service can guarantee absolute security.

When processing or providers change, we will update this policy and indicate the update date.

The legal review must consider the entry into force of Law 21.719, scheduled for 1 December 2026.