akhmadjonovsherzod/multisource-retail ? reverse-engineered prompt

Build me an end to end retail data pipeline that generates a realistic synthetic dataset, then processes it with PySpark into a clean order level fact table.

I want it to take customers, products, orders, and order items from different file formats, combine them, filter out cancelled or returned orders, and calculate customer analytics like running spend, order rank, spend rank, and days since the previous order. Save the output as partitioned Parquet by month.

Then upload that data to S3, load it into Snowflake, and transform it with dbt into a few useful reporting tables like monthly performance and customer summaries. Add schema tests so the pipeline checks for bad or missing data.

Finally, wire the whole thing together in Airflow with a daily DAG so the steps run in order automatically. Use environment variables for credentials, keep it Docker friendly, and if you need to, look up current docs online for PySpark, Snowflake, dbt, or Airflow setup details.