nhl-bigquery 0.2.0

NHL play-by-play → BigQuery: idempotent ingestion + LLM-friendly docs + verification

nhl-bigquery

Idempotent NHL play-by-play (with on-ice arrays merged from shift-charts) → BigQuery ingestion, with first-class documentation for SQL/LLM agents and verification against the NHL public API.

Install

pip install nhl-bigquery

Quickstart

gcloud auth application-default login
nhl-bigquery sync \
    --start 2024-10-01 --end 2025-06-30 \
    --plays-table myproject.mydataset.nhl_plays

This writes six tables to myproject.mydataset.*:

• nhl_plays — one row per event, with home_on_ice_ids / away_on_ice_ids arrays
• games — schedule dimension
• game_officials — referees + linesmen per game
• boxscore_stats — per-player per-game stats
• shifts — per-shift per-player intervals
• standings — daily team-standings snapshots

Backfill

Backfill 15 seasons in resumable monthly chunks:

nhl-bigquery sync \
    --start 2010-10-01 --end 2026-05-11 \
    --chunk-by month --resume \
    --plays-table myproject.mydataset.nhl_plays

--resume skips chunks already recorded as success or empty in <dataset>._nhl_ingest_runs. Re-running with the same --chunk-by is safe; switching between runs will re-process (chunks must match exactly).

Documentation

nhl-bigquery docs --format llm > NHL_FOR_LLMS.md
nhl-bigquery docs --format bq-apply --table myproject.mydataset.nhl_plays

Five formats: bq-apply (push descriptions to BigQuery), llm (one Markdown file packing every column for LLM context), dictionary (JSON rows for a data dictionary table), markdown (human reference), and dbt (dbt YAML schema stub).

Verification

nhl-bigquery verify --source internal \
    --aggregation internal-consistency \
    --table myproject.mydataset.nhl_plays

nhl-bigquery verify --source nhl-api \
    --aggregation team-season --metric all --season 2024 \
    --table myproject.mydataset.nhl_plays

MIT licensed. This software does not include or distribute NHL data.