README.md

---
license: mit
sdk: gradio
colorFrom: yellow
colorTo: green
sdk_version: 6.0.2
---
# Sundew Health Backend

Neurosymbolic, energy-aware ECG monitoring backend using FastAPI, PyTorch, and PostgreSQL.

## Quickstart

1. Create a `.env` from `.env.example` and set `DATABASE_URL`.
2. Install dependencies (CPU PyTorch wheels): `pip install -e . --extra-index-url https://download.pytorch.org/whl/cpu`.
3. Run the API: `uvicorn app.main:app --reload`.

## Database (PostgreSQL + Alembic)

- Set `DATABASE_URL` to your Postgres DSN (e.g., `postgresql+asyncpg://user:pass@localhost:5432/sundew_health`).
- Run migrations: `alembic upgrade head`.
- Create new migrations: `alembic revision -m "message" --autogenerate`.

## Tests

Run `pytest` to execute the test suite.

## Notes

- Gating uses `sundew-algorithms` (significance + hysteresis) ahead of model inference.
- Adaptive Sparse Training (`adaptive-sparse-training`) is installed; torch shims are applied to load it, but training still uses the simpler loop until AST wiring is added.