An agency to enable ad-hoc english queries against the data, plus some views which contain pre-calculated KPI values based on the dataset for queries about performance against those KPIs.
Because this is completely experimental it is currently implemented in a Jupyter notebook.
Check the Agency Swarm repo for more information on the underlying framework and how to use it.
Download from https://www.kaggle.com/code/ahmedessamsaber/coffee-shop-sales a single file containing coffee shop transactions. Import this into your database, the code currently expects an MS SQL Server db but this would be simple to change.
The table name to import into should be CoffeeShopSales
Currently only 1 view, use the sales_analysis.sql script to add the view to your db.
This code assumes an MS SQL Server but you could easily implement any database in the SQL Ad-hoc Query Tool.
This mini-project was done experimentally, as such an ad-hoc tool is useful enough but for a real agency I would likely prefer more specific tools for specific queries where that was applicable at least. This will increase consistency & reliability while reducing hallucinations.
You will need to add a file config.json in the same directory which should contain your MS SQL db connection string like:
{
"connection_string" : "mssql+pymssql://<user>:<password>@<yourdb-server-url>/<yourdb>"
}