SleepSync

Proof-of-concept for scheduling an overnight melatonin patch profile. Expo app, FastAPI backend, optional ESP32 over BLE.

You set bed and wake times. The backend returns a dose curve for that window. The app runs it overnight on a BLE patch or in an on-screen simulator if you do not have hardware. A short morning debrief (did you wake, how groggy) closes the loop and shapes the next plan.

Inputs. Mock sleep data by default. Connect Google Health to pull real staging and vitals.

Outputs. A piecewise profile (delay, ramp, sustain, taper). Dose is normalized [0, 1], not mg. Wake risk comes from a hand-tuned heuristic (heuristic-v0), not a trained clinical model.

Development

docs/RUNNING.md

Further reading

• docs/GOOGLE_HEALTH.md optional wearable sleep input
• docs/BLE_PATCH.md ESP32 BLE protocol
• backend/README.md API and plan pipeline

Layout

mobile/     Expo app
backend/    FastAPI, optimizer, SQLAlchemy
firmware/   ESP32 NimBLE sketch
docs/
shared/     test fixtures

Team

Charles Muehlberger (@charlespers), Phu Duong (@phuhduong), Jaime Nunez (@Jaimenunez10), Tom Wang (@tom05919)