Skip to content

README.md

Latest commit

 

History

History
56 lines (42 loc) · 4.08 KB

README.md

File metadata and controls

56 lines (42 loc) · 4.08 KB
Raw

LifeOS

An open-source, self-hosted, AI-native operating system for your body and day.

LifeOS records the things most apps scatter across a dozen silos — sleep, food, skincare & haircare, body/face/nail progress photos, gym, wardrobe, books, mood, supplements, egestion, and more — onto one unified timeline, then uses AI to surface the cross-domain cause-and-effect that fragmented trackers can't.

Why another tracker? Because no app connects what you eat / how you sleep / which products you use to how your skin, gut, body, and mood actually respond. LifeOS is built around that missing link — and it's yours: self-hosted, your data, your keys.

status license PRs welcome PWA backend

⚠️ Early & building in public. The architecture, decisions, and scaffold are in place; the domains are being built out. ⭐ Star/watch to follow along.

🔴 Live demo: lifeos web app → · API + interactive docs: /docs

✨ What it tracks

  • 😴 Sleep · 🍳 Nutrition & hydration (+ caffeine/alcohol) · 🚿 Skincare / hair / body care with product + routine tracking and ingredient-conflict awareness
  • 📸 Face / skin / hair / nail progress photos (alignment overlays) · ✋ nail-biting & habit tracking
  • 🚽 Egestion (Bristol + urine scales) with doctor-shareable reports · 🏋️ Gym · 👕 Wardrobe & outfits · 📚 Books (in-browser EPUB reader)
  • 🧠 Mood / energy / stress · 💊 Supplements & meds · 🌦️ auto weather + location context
  • 📈 a daily timeline + cross-domain insights — built to be honest: confidence-scored, never overclaiming, never medical advice

🧱 Architecture at a glance

  • Timeline-spine data model — an append-only event log over normalized per-domain tables, so every domain shares one searchable, embeddable history.
  • Web-first PWA — React 19 + Vite + Tailwind v4 with an offline-first capture queue (log anywhere, sync later).
  • Python backend — FastAPI + DSPy, async SQLAlchemy on PostgreSQL + pgvector (one datastore), Procrastinate for background jobs (no Redis).
  • AI-native, model-routed — all LLMs via OpenRouter (Claude + Gemini), Voyage embeddings, ElevenLabs voice; image cutouts via Gemini "nano banana"; deterministic libraries for color.
  • Private by construction — self-hosted on Railway + Cloudflare R2, app-side AES-256 on sensitive media, zero-retention AI routing, one-click full export.

📖 Full design in docs/ARCHITECTURE.md, the decision log in docs/DECISIONS.md, and the grounded research behind every choice in docs/research/.

🚀 Getting started (dev)

# backend (needs PostgreSQL with the `vector` extension)
cd api && cp .env.example .env      # fill in your keys
uv sync && uv run alembic upgrade head && uv run fastapi dev app/main.py

# frontend
cd web && pnpm install && pnpm dev

Integration setup guides: Cloudflare R2 · Withings.

🗺️ Roadmap

Phased build — see docs/PHASES.md. Now: foundation + first domains (sleep · nutrition · mood · egestion · care) + the timeline. Next: books, wardrobe, gym, the photo + insight pipelines.

🤝 Contributing

Early days — issues, ideas, and PRs are welcome. Start with docs/ARCHITECTURE.md and CLAUDE.md.

🔐 Privacy

Single-user and self-hosted: your data lives in your Postgres and your object storage. Sensitive media is encrypted with a key only you hold. No analytics, no third-party data sale.

📄 License

MIT © 2026 Prince Upadhyay