I'm building a nutrition planning app for endurance athletes (Getfed) and want to confirm an architecture is compliant with Section 5.3 before applying to the Developer Program.
Section 5.3 prohibits using Strava Data in connection with the operation of an AI Application, including ingestion into a context window or working memory.
My data flow keeps Strava Data out of any model entirely:
Strava API → deterministic service → macro targets → LLM → meal suggestion
• The deterministic service reads activity duration, intensity and estimated expenditure via the API.
• It computes daily protein/carb/fibre targets using published formulas (Mifflin-St Jeor, ISSN/ACSM position stands). No model involved — pure arithmetic.
• The LLM receives only the resulting macro targets (three numbers) plus stored food preferences. No Strava Data, no activity metadata, nothing else derived from Strava.
• Strava Data is never used to train, fine-tune or evaluate any model.
• Data is shown only to the athlete it belongs to.
Questions:
1. Does this fall outside Section 5.3 as written?
2. Is there an approved pattern for apps that use Strava Data deterministically while using AI elsewhere in the product?
3. Anything specific I should document for the Developer Program application?
Happy to share an architecture diagram if useful. Thanks!
