1. Photo analysis and image generation
Reading a photo you select is different from generating an illustration. Food estimates and plan transcription analyse an input image; meal and exercise illustrations are synthetic artwork. A generated dish is not a photograph of your logged meal, and a generated athlete is not a photograph of you or proof of a real person’s results.
2. Where information goes
The following reflects the AI routes verified on 15 September 2026. Model authors and the service that receives a request are different roles. The current Cloudflare routes are hosted by Cloudflare; the reviewed implementation does not call Black Forest Labs directly.
| Feature | Information sent | Processing services |
|---|---|---|
| Intelligence and AI briefings | Messages and the relevant profile, training, nutrition, body and recovery context; memory where enabled. | Google Gemini. |
| Workout and nutrition plan generation; text/file parsing | Your request, selected plan/document text, goals and relevant preferences. | Google Gemini. |
| Food-photo analysis | The selected image and the instructions/language needed to estimate its contents. | Cloudflare Workers AI using Google’s Gemma model. This is not a call to Google Gemini simply because Google developed the model. |
| Workout-sheet photos and nutrition-plan images | Selected pages or photos, potentially including names, annotations or health information visible in them. | Cloudflare Workers AI/Gemma; extracted nutrition-plan text is subsequently processed by Google Gemini. |
| Meal illustrations | Meal and option labels, ingredient descriptions, portions and language. This generation path sends text, not your original food photo. | Cloudflare Workers AI with Black Forest Labs FLUX.2 [klein] 4B; an eligible failure can use the implemented Gemini image fallback. |
| Exercise illustrations | Exercise names, equipment and muscle descriptions. Exercise-name classification uses a separate request. | Google Gemini for classification, Cloudflare Workers AI/FLUX.2 for generated artwork, with an implemented Gemini image fallback. |
3. Choose uploads and labels carefully
Only upload documents, photos or other content you are entitled to use for the requested processing. Crop unrelated names, contact details and medical information when they are not needed. Do not include another person’s private information without an appropriate legal basis.
Meal, option and custom-exercise labels can themselves contain personal information and can enter prompts or catalogue metadata. They are not automatically anonymous because an account identifier is omitted. 
remains responsible for appropriate minimisation and safeguards; these instructions do not transfer that responsibility to you.
4. Shared illustrations and your private files
Equivalent meal or exercise requests can reuse a generated illustration from a shared asset cache or exercise catalogue. The same generated artwork may therefore appear for more than one user. This reuse is an app storage feature; it is not the training of the underlying AI models.
The reviewed generation path stores generated output in shared asset paths. It does not copy your original uploaded food/body photos into that shared library. Account-linked originals and personal records follow the privacy and deletion procedures.
A generic, non-personal shared illustration can remain after an account is deleted because it is used by the catalogue. This is not an exception allowing personal information to be retained without a lawful basis. Contact support if shared artwork or associated metadata identifies you or contains private information.
5. Retention and provider settings
stores returned images and references for the requested feature or reusable catalogue. Provider-side processing and operational logs are separate from those stored copies. The service-provider register explains international processing and the applicable provider notices. No promise of zero retention at every provider or of instant deletion from all backups is made.
6. Rights, permitted use and limits
You retain the rights you have in material you upload. You authorise processing necessary for the feature you request and the sharing you choose, subject to the privacy policy. A model’s commercial licence does not guarantee that every output is copyright-protected, exclusive, factually correct or free of another person’s rights.
FLUX.2 [klein] 4B and the reviewed Gemma 4 model are made available under Apache 2.0 terms. The hosted service also has its own terms. does not distribute these model weights to app users. See Black Forest Labs licensing, Gemma licence and Cloudflare data/model notice.
Do not use generated media to impersonate a real person, falsely suggest endorsement, misrepresent body transformations or distribute unlawful material. Model licences do not grant rights to third-party trademarks, likenesses or source documents.
7. Fitness and transparency
AI can misrepresent anatomy, apparatus, posture, ingredient identity or portions. Do not treat an illustration as validated exercise instruction, an allergen guarantee or proof of a nutritional value. Verify important details independently.
The AI transparency notice describes the service and the European AI Act. Visible disclosure and applicable machine-readable provenance obligations must be addressed in the actual app and media-delivery/export flows; this page alone does not certify them.
8. Report an image or rights concern
Contact strenqo-support@strenqo.eu with the feature, exercise/meal identifier or enough detail to find the image, and the reason for your concern. For a rights claim, explain your relationship to the work or person and the requested correction/removal. Include only necessary evidence, not an entire private health file.