Skip to content

AI analysis

Your data, your model, your questions.

Bring your own API key for Claude, OpenAI, DeepSeek, Qwen or a local Ollama model. MoveVitals sends it a clean, source-tagged summary of the data you allow — and nothing else.

01

Bring your own key

Use the provider you trust. Keys are encrypted with AES-256-GCM and only used when you ask a question.

02

You choose what's shared

Pick which sources the model may see. Data whose platform terms restrict AI use is excluded by default.

03

Grounded in your numbers

Answers draw on resolved daily metrics, workouts and your body profile, with sources and conflicts marked.

04

Never used for training

Your data is never used to train models — by us, and only per your provider's own terms when you send it.

Questions people ask

  • Summarise my recovery over the last two weeks — am I ready for a hard session today?
  • Why did my HRV drop this week, and does it line up with my training load?
  • My WHOOP and Garmin disagree on resting heart rate. Which should I trust?

Questions about your data

What data is sent to my AI provider?
The analysis context contains source-tagged daily metrics and workouts, normally over 42 days. Profile fields, lab results and workout names follow your settings; lab inclusion is off by default. Strava is excluded by default unless you explicitly allow it. Check the sources and privacy options before running analysis.
Does self hosting keep all AI processing local?
Only if you configure a local model endpoint. A cloud endpoint receives the permitted context. Local/private endpoints require the self-hosted operator to enable private-URL access; the selected provider's processing and retention policies apply.
Who pays for the model, and can its advice be wrong?
You supply your own AI API key and pay any model-provider charges. AI output may be wrong or incomplete and is not a diagnosis or treatment plan. The key is distinct from wearable-account authorization.