Most work on Off The Pace never needs a full backfill. Pick the smallest scope that fits the task you can always pull more later, since every ingest.py run is idempotent and skips what’s already on disk unless you pass --force.
Test fixtures
Single race
Recent seasons
Full backfill
Offline tests
Use when: writing or testing dbt SQL transforms.No network call at all transform/tests/fixtures/bronze/ ships a small, committed slice of three real races (a low-energy circuit, a clean dry race, a wet/mixed race) that dbt can build against directly via the bronze_base var.This is the default starting point for almost any transform-layer change it’s also exactly what CI runs.
Use when: smoke-testing ingestion itself, or you only need one specific session.Roughly 2–5 minutes and ~100 MB, derived from ingest.py --dry-run’s own estimate (races × 0.15 GB). Use when: scale-testing a transform change or validating it before a PR recent seasons are enough to exercise current schema shapes without waiting on a full backfill.~15 minutes, ~200 MB. Use when: training ML models, running full verification, or anything that depends on the complete historical record.A full backfill takes 30–45 minutes and roughly 2 GB on disk. It’s the slowest path confirm you actually need every season before reaching for it.
Use when: developing ingestion itself no network, no Bronze write.Exercises DataQualityEngine and the ingest pipeline against small in-memory DataFrames and mocked FastF1 responses no Parquet fixtures, no real API call. Mix and match with flags
Every scope above is a starting point, not a fixed menu combine ingest.py’s flags (see CLI) to narrow further: --session R to skip qualifying, --skip-telemetry for a faster pull when you don’t need it, or --round N to pin a single race within a season.
Re-running any of these commands is safe. ingest.py skips a session that already exists on disk unless you pass --force so reaching for make ingest-all after already having make ingest-recent only pulls the seasons you’re missing.
