Why I built an open source alternative to Terra for wearable data

说明与用途

作者创建了 Open Health Sync，一款自托管的可穿戴设备数据聚合器，保留原始 payload，支持 FHIR R4，并在低资源设备上运行，提供比付费托管 API 更经济的方案。

问题

Existing hosted APIs like Terra are expensive, introduce third‑party data paths, lose raw payloads, and do not enforce FHIR R4 standards, making them unsuitable for projects that need cost control, full data retention, and clinical interoperability.

结果

Open Health Sync achieves 2–3× faster end‑to‑end ingestion than Open Wearables, uses ~22 MB idle memory versus ~494 MB, and stores raw payloads for unlimited retention while providing FHIR R4 and other export formats.

工作流

• Clone the repository and set up environment variables
• Generate an encryption key and start the Docker Compose stack
• Run the health‑check command
• Seed synthetic data for demo or import real Apple Health export
• Use the REST API or FHIR endpoints to query merged data
• Delete user data via DELETE /v1/users/{id} if needed

证据

Every wearable speaks its own dialect. Apple Health sends one JSON line per HealthKit sample, with heart rate in count/min and oxygen saturation as a fraction. Withings sends a daily activity object keyed by date and intraday series keyed by Unix epoch. Oura, WHOOP and Fitbit each have their own idea of what a day, a device and a duplicate are.

来源

It idles at about 22 MB of memory and runs on a Raspberry Pi 5.

来源

Open Health Sync stored the records 2 to 3 times faster end to end and used about 22 MB of memory idle against about 494 MB, while Open Wearables used 3.4 times less disk.

来源