Dataset plugins — shenas open source

What a dataset is

A dataset is a versioned table schema, declared as a Python dataclass extending DatasetTable. It specifies the column names, types, units, and primary key for a canonical metric — for example,daily_hrv, transaction, or activity.

Multiple source-transform pairs can write into the same dataset table. A fitness-tracker source + a geofence transform and a phone-health source + a SQL transform can both write activity rows. Surfaces then read a unified, schema-stable table regardless of origin.

Schema stabilityDataset schemas are versioned. Breaking changes require a new major version. Transforms and surfaces pin to a schema version.

UnitsColumn names carry unit suffixes (duration_s, distance_m, weight_kg). All units are SI unless the upstream API forces otherwise.

Entry pointRegistered under shenas.datasets. Install with shenasctl source add --local; shenas creates the table on first transform run.

Community datasets

The canonical community datasets (activity, daily_hrv, transaction, sleep, event, note) are maintained under plugins/datasets/in the main repository: shenas-org/shenas Public mirror (shenas-org/shenas): Last updated 2026-05-23. Treat it as frozen; do not rely on it for current releases. If a dataset you need doesn't exist, open a PR to propose the schema — the bar is stability and SI units.

activityWorkout or movement session: duration_s, distance_m, sport type, start_at, device_id.

daily_hrvDaily heart rate variability: date, rmssd_ms, sdnn_ms, source.

transactionFinancial transaction: date, amount, currency, merchant, category, account.

sleepSleep session: start_at, end_at, total_s, deep_s, rem_s, device_id.

eventCalendar event: title, start_at, end_at, calendar, location.

This list is illustrative; authoritative definitions are in the repo. Dataset schemas are still being finalized — check theroadmap for stability status.