# PULS3

PULS3 is a privacy-first AI health coaching product. It turns Apple Health and wearable signals into practical daily actions across sleep, nutrition, exercise, stress, and menopause support.

## Architecture

PULS3 uses a multi-agent system called Consilium. A Coach agent orchestrates specialist agents (Sleep, Nutrition, Exercise, Stress, Biomarker, Women's Health, Menopause, Men's Health), each with its own memory namespace, clinical screening instruments, and domain knowledge. The model is interchangeable — the harness (safety engine, structured memory, agent orchestration, clinical governance) is the product.

## Clinical Foundations

Each specialist agent uses validated screening instruments with published scoring thresholds:
- Sleep: Insomnia Severity Index (ISI), CBT-I protocol
- Nutrition: AUDIT-C alcohol screen, Mediterranean Diet Score, NHMRC Australian Dietary Guidelines
- Exercise: PAR-Q+ readiness screen, FITT-VP framework, Australian Physical Activity Guidelines
- Stress: PSS-10, PHQ-2/PHQ-9 depression screen, GAD-7 anxiety screen
- Women's Health: cycle-phase-aware coaching, iron/ferritin awareness, PCOS and endometriosis lifestyle strategies
- Menopause: AMS Symptom Score Card, STRAW+10 staging, CBT for menopausal symptoms
- Men's Health: andropause coaching, PSA literacy, ASCVD lifestyle-first prevention, cortisol-testosterone interplay
- Biomarker: RCPA reference ranges, ASCVD risk estimation, clinically significant deltas

All agents have explicit escalation thresholds for recommending professional help.

## Key Capabilities

- Lab report ingestion: upload PDFs or photos in chat, parsed with three-tier fallback (text extraction, spatial OCR, vision)
- Meal and form-check photo intake with on-device face and sensitive-text redaction
- Proactive follow-ups that are consent-first and engagement-gated
- Life-stage-aware coaching adapts for menopause, andropause, parenthood, and shift work
- Outcome tracking with baseline and follow-up scores on validated instruments

## Privacy

- On-device storage by default (health data, conversations, memory, safety logs)
- Cloud LLM inference required for coaching responses — transient, nothing stored server-side
- All image processing happens on-device — no unredacted images leave the phone
- Pseudonymous cloud model processing where needed
- Prompt sanitization and no-retention controls where supported
- Safety-focused coaching, not diagnosis or treatment

## Canonical Site
- https://puls3.app/

## Primary Pages
- Home: https://puls3.app/
- Sleep coaching: https://puls3.app/sleep
- Nutrition coaching: https://puls3.app/nutrition
- Exercise coaching: https://puls3.app/exercise
- Stress coaching: https://puls3.app/stress
- Women's health coaching: https://puls3.app/womens-health
- Menopause coaching: https://puls3.app/menopause
- Men's health coaching: https://puls3.app/mens-health

## Policies
- Privacy: https://puls3.app/privacy
- Terms: https://puls3.app/terms