Vita is a personal health companion that uses Gemini 3 as its core intelligence layer across three distinct integration surfaces.

Agentic Health Assistant. Each user receives a dedicated Gemini-powered agent with an isolated file-based workspace. The agent reads from nine canonical health data stores: HealthKit vitals, sleep sessions, workouts, nutrition logs, hydration records, FHIR medical records, wellness plans, user goals, and profile data that materialized as structured markdown context for every agent run. This grounding enables deeply personalized, evidence-based health guidance rather than generic responses. The agent maintains long-term memory across sessions through workspace files it reads and updates autonomously.

Multimodal Analysis. Vita calls the Gemini API directly from the iOS client for two vision tasks: (1) meal photo analysis, where Gemini identifies foods, estimates macronutrients, and returns structured JSON; and (2) medical document interpretation, where Gemini performs OCR on lab results, prescriptions, and imaging reports, extracting key entities and clinical insights.

Structured Output Generation. Beyond conversational chat, Gemini generates JSON-schema-constrained outputs for daily health insights synthesized from recent activity data, and multi-week wellness plans with exercise schedules, nutrition guidelines, and sleep recommendations, all personalized to the user's goals and health history.

Gemini 3 is not just an add-on; it is the reasoning engine that connects raw health data to actionable guidance.

Share this project:

Updates