app.ingest.replay

Replay source: plays traffic.csv into the ingest pipeline as the dataset clock passes each row.

On start it backfills the last backfill_s of dataset time at once — the prediction features need recent GPS history (about an hour), and a live feed alone would leave the model blind until it builds up.

Classes

ReplaySource(pings, clock, accept, *[, ...])

class app.ingest.replay.ReplaySource(pings: Sequence[Ping], clock: DatasetClock, accept: Callable[[Ping], bool], *, backfill_s: float = 3600.0, tick_s: float = 1.0)

Базовые классы: object

__init__(pings: Sequence[Ping], clock: DatasetClock, accept: Callable[[Ping], bool], *, backfill_s: float = 3600.0, tick_s: float = 1.0) → None

async run() → None

snapshot() → dict