Monthly data quality report for taxi trips

#Brief

Every month the taxi data arrives, and every month someone asks "can we trust it?". Automate the answer.

#What to build

1. A script quality.py --month 2025-01 that runs at least ten checks with DuckDB directly on the TLC Parquet file.

2. Measure normal rates over several months (negative fares, missing passenger counts, pickups outside the month) and set warn and fail thresholds from them.

3. Include the consistency check between total_amount and its parts, and report the pattern of differences.

4. Output a Markdown report with a verdict, a table of checks, and month-on-month changes.

#Deliverables

• quality.py, three reports, escalations.md

#How it is graded

Reviewers run the script on a month you did not include. The report must work unchanged and give a sensible verdict.

#Stretch

Run it as a GitHub Actions job on the 5th of every month and open an issue when the verdict is fail.