Hotfix: address missing data

[jairus-m]

Summary

Production nightly runs have been successful so far. I added logging early on to output metadata throughout every step in the pipeline; however, the metadata aren't tied to asset checks outside of dbt. I checked the logs, and saw that the size of the cycling data from model training was less than 600 activities (in the cycling_data asset). Since this is my personal data, I knew to expect +1000 activities so the data is obviously wrong. This PR addresses this issue.

Note: This also prompts a future feature need for backfilling capabilities in the dltHub pipeline in the context of dagster.

Details

Source of the issue

• incorrect duckdb schema config in __src_schema.yml

Additions:

• minor updates to dbt models / duckdb query

• add custom dbt tests in analytics_dbt/tests/

  • analytics_dbt/tests/test_activity_counts.sql
  • analytics_dbt/tests/test_cycling_counts.sql

• add dagster asset_checks for cycling_data

  @asset_check(asset=cycling_data)
  def check_cycling_data_size(cycling_data):
      """Check that the cycling data has more than 1500 rows."""
      num_rows = cycling_data.shape[0]
      passed = num_rows > 1500
      return AssetCheckResult(
          passed=passed,
          metadata={"num_rows": num_rows}
      )