Privacy policy

1. What Soma does

Soma is an iPhone app for adults aged 18 and older to track food, water, protein, calories, and steps. There is no Soma account or journal sync service. We use information to provide logging, history, exports, optional personalized AI answers, subscription management, support, and service security.

We do not sell your personal information, run advertising in Soma, or use your Health data for advertising, marketing, or unrelated data mining. Nutrition estimates are general information, can be inaccurate, and are not medical advice.

2. Information stored on your iPhone

Soma stores the profile and goals you enter, including an optional name and profile photo, age, sex, weight, height, activity level, measurement preferences, and daily targets; food, water, steps, and history records; and AI journal entries, including selected images and saved voice memos. These support your progress, history, editing, and exports.

The app uses local iOS storage and files. Its local profile, history, journal photo, and voice files are excluded from automatic device backups because they contain personal health information. Copies in older backups or exports you control are separate. Soma has no account backend holding a synchronized copy of your journal. Local storage does not mean information included in an authorized AI request stays on the device.

3. Apple Health and HealthKit

With the Health permissions you choose, Soma reads step count to show activity and writes dietary energy, protein, carbohydrates, fat, and water from your logged food and drink. Separate weight and height controls let you explicitly import the latest readable body mass and height into your profile after reviewing the measurements and their dates.

When authorized, confirming a weight or height edit in Edit Profile can write that measurement to Apple Health. Imported measurements are not written back as new samples. Details supplied in chat are saved locally; chat does not itself write body measurements to Health. Apple Health holds its own copies under your Apple settings, separately from Soma’s local journal.

Health access and cloud sharing are separate choices. Granting Health access does not by itself authorize sharing with Google. When you separately allow cloud AI, saved Health-derived step totals and weight or height you have imported into your profile may be included in the chat context sent through Cloudflare to Google Gemini for nutrition and personal progress questions. This is selected context, not access by Google to your entire Health database.

We do not send Health information for advertising, marketing, or general-purpose model training. Google’s Paid Services data-processing terms apply to AI content; its security and abuse-monitoring retention still applies. Health information is not sent to RevenueCat for subscription management.

You can change Soma’s Health permissions in the Health app or iPhone Settings. Withdrawing access stops the related reads and writes going forward; manual logging remains available. It does not erase previously imported profile values, previously written Health samples, or content already sent to an AI provider. Reset All Data attempts to remove only samples created by Soma, including its body measurements; it does not delete other apps’ Health data. You can review or remove remaining samples in the Health app.

4. Google Gemini: data shared and purpose

Soma uses Google’s Gemini API, accessed through a Google Cloud project with billing enabled, as its external AI processor. Requests pass through Soma’s proxy hosted on Cloudflare Workers. Google processes the request content to interpret meals, transcribe recordings, estimate nutrition, answer questions, and return results. Soma does not host or train the Gemini model.

Your saved profile name, profile photo, and full library of journal photos or recordings are not automatically attached as chat context. Messages, selected images, or recordings may nevertheless contain personal information you choose to include. Missing history is not treated as proof that you ate or drank nothing.

Voice, photos, and on-device features

When voice audio is uploaded, Soma sends the recording to Google without the device recognizer’s transcript. Google transcribes that audio, and a subsequent request analyzes those model words. An explicitly selected transcript-only route may send recognized speech text instead. A failure to prepare an audio upload stops that attempt rather than silently substituting the device transcript. Apple speech recognition may also process audio through Apple’s services depending on device support and Apple settings.

Photos normally use Google Gemini. Supported on-device photo analysis can be tried first, but a cloud fallback remains possible if you have allowed cloud AI. Text estimates can start on-device and receive a cloud nutrition check. Conversational chat uses the cloud; local daily status answers remain available. An on-device-first setting is not a guarantee that a request will stay on the iPhone.

Google’s use and retention

Under the Gemini API Paid Services terms, Google does not use submitted prompts, files, or responses to improve its products. It processes this content under its data-processing terms. This is the developer API with billing enabled, rather than the consumer Gemini chat app or an unpaid API project.

Google’s abuse-monitoring documentation states that it retains prompts, additional context, and generated responses for 55 days to detect and prevent policy violations, maintain service safety and security, and meet required legal or regulatory disclosures. Flagged content can be reviewed by authorized Google personnel. Abuse-monitoring data is used for policy enforcement, including models used specifically for that enforcement, rather than general-purpose model improvement. We do not promise zero retention or the absence of human review.

Google also processes service usage and technical information needed to operate and secure the API, such as token counts, operational status, safety-filter events, authentication information, and network metadata, under its applicable terms. Google receives the connection from our Cloudflare proxy; the Worker does not deliberately forward your iPhone’s source IP as part of the AI content. Cloudflare receives the original app connection, as explained below.

Your permission and choices

Before a cloud AI request, Soma requires an explicit cloud-sharing decision and separate adult confirmation. The permission screen describes the content and purpose. Earlier grants are reviewed again when the app expands the data included, such as saved history or optional profile details. This policy identifies Google Gemini and Cloudflare as the recipients.

Choosing Not now, or leaving the permission screen without allowing sharing, blocks cloud requests, including background nutrition checks and cloud fallbacks. You can review the decision from the AI tab or turn off Cloud AI sharing in Settings → AI & Siri. Manual logging, local daily status answers, Siri water logging, and supported on-device analysis remain available. Siri meal logging requires prior cloud permission.

Turning cloud sharing off stops new AI requests; it cannot withdraw a request already processed or erase providers’ retained copies. Camera, microphone, speech, location, and Health permissions are additional iOS choices, not substitutes for cloud AI permission.

5. Cloudflare Workers and Apple App Attest

Cloudflare hosts the proxy that receives and forwards authorized AI content to Google and returns the response. It processes request bodies while handling them and connection information such as IP address, request URL, time, and technical service metadata to deliver and secure the service. The Worker’s application code does not persist or print meal descriptions, Health/profile content, photos, recordings, chat bodies, or Gemini responses.

Apple App Attest helps verify that cloud requests come from a genuine Soma installation and prevents forged or replayed requests. Apple processes key-attestation requests. The proxy receives and verifies installation-key identifiers, public keys, Apple attestation evidence/receipts, app and environment information, and signed request assertions. It retains the public-key record, registration evidence, app/environment metadata, and request counter in Cloudflare Durable Objects for ongoing verification and replay prevention. These records identify an installation; they are not your name or an advertising identifier.

The Worker hashes the connecting IP for rate limiting. A hashed IP is still security-related data; we do not describe it as anonymous. Registration challenges expire after three minutes and request challenges after one minute. Rate-limit windows are normally one minute, with a one-hour registration limit. Expired challenge and rate-limit records are removed during later challenge processing.

Installation-key records currently have no automatic expiry. Resetting local app data, turning AI off, or uninstalling does not automatically delete those server records. See retention and deletion for how to request review of information we can identify and control.

Cloudflare’s operational logging is enabled. It can retain invocation metadata and fixed diagnostic reason codes; the Worker does not deliberately log private content. Cloudflare currently documents standard Workers Logs retention of three days on its Free plan or seven days on its Paid plan. Other Cloudflare service/security records follow its applicable service terms. See Workers Logs retention and Cloudflare’s privacy policy.

AI transport uses HTTPS. App Attest verifies requests; it does not hide content from Cloudflare or Google when they process it and does not mean that cloud AI has no retention.

6. Apple, RevenueCat, and other device services

Apple’s platform services handle their own information under Apple’s privacy policy. A subscription or security identifier can exist even though Soma has no account or journal sync service.

7. This website and support email

This static website is hosted by Vercel. Its files contain no analytics script, ad tracker, contact form, or site-set cookie. Vercel processes visitor request details, including IP address, browser/system information, requested page, and access time, to deliver and secure the site under its privacy notice. Visiting this site does not submit your Soma journal to Google.

If you email us at our Gmail address, we and our email provider, Google Gmail, receive your email address and the message and attachments you choose to send. We use the exchange for support and privacy requests, and retain it as needed to resolve the request and meet legal obligations. This support email processing is separate from the Gemini API’s 55-day abuse-monitoring period. See Google’s privacy policy. Avoid sending sensitive health details unless necessary.

8. Retention, deletion, and your choices

Reset All Data preserves Apple-managed subscription status and the consumed AI scan balance. Follow Data choices to export or reset Soma, revoke cloud permission, manage Health access, and cancel Pro through your Apple Account.

Depending on where you live, you may have rights to request access, correction, deletion, portability, or restriction of personal information we control. Email us for these requests, including identifiable proxy security or RevenueCat records. We will explain what we can identify and control, any verification needed, and applicable legal limits. There is no Soma account, and we may be unable to match an installation key or provider log to an email address alone. Do not send private signing keys, security assertions, or your full health history.

We cannot promise immediate deletion of all provider logs or records that must be retained for security or legal reasons. We use providers under their applicable service and data-processing terms. Providers may process information outside your country; Google’s terms allow processing in countries where it or its agents maintain facilities. We will update this page and its effective date when our practices materially change.

9. Contact

For privacy questions or requests, contact Zeyad Hussein (Hyp4tia) at ziadm4772@gmail.com. Describe your request and avoid sending sensitive health information unless necessary. We may need enough detail to identify a support exchange or provider record, but there is no Soma account to look up.