grig-teo.spaceIn development
A monorepo that powers my public site and a complete personal health pipeline. A SwiftUI app reads my COLMI R10 ring over a reverse-engineered BLE protocol with no vendor cloud, and streams heart rate, SpO2, stress, HRV, sleep stages and activity into a NestJS/Postgres backend. That backend turns the data into hourly AI tips, a recovery score, streaks, weekly digests and an AI doctor, which queries any metric over any date range, reads scanned lab results and examines note photos through a local vision model. It ships as a Next.js site (en/ru/ro) with an admin dashboard, an iOS app with a widget and HealthKit gap-fill, a macOS tip app with notifications, and a Telegraf bot, all self-hosted on a single VPS.
Highlights
- — Reverse-engineered the COLMI R10 BLE protocol in Swift: heart rate, SpO2, stress, HRV, sleep stages and activity sync straight to my server, with no vendor cloud
- — Built an AI doctor (Kimi K3) with tool access to the whole health database: any metric over any date range, sleep stages, activities, notes and document scans, including images
- — Generated hourly AI health tips from live metrics, weather correlations and journal notes, using DeepSeek with an on-VPS Ollama fallback
- — Built the sleep pipeline: stage parsing, composite scoring, a recovery score against 14-day HRV and resting-HR baselines, and deviation alerts
- — Ran four surfaces on one backend: a Next.js 15 site (en/ru/ro) with an admin dashboard, a SwiftUI iOS app with a widget, a macOS tip app and a Telegraf bot with photo and video notes
- — Self-hosted the entire stack on a single VPS with no third-party cloud: Postgres, MinIO with public and private buckets, an Ollama vision model, Docker and nginx
Stack
TypeScript · Next.js 15 · NestJS · SwiftUI · CoreBluetooth · HealthKit · Telegraf · PostgreSQL · TypeORM · MinIO · Docker · nginx · Kimi K3 · DeepSeek · Ollama