You finished the schema. The models are in, the first screen renders. Then you go to build the real thing and there is nothing there. The table is empty, and you cannot test a feature against zero rows.
So you improvise. A few hand-written fixtures. A prompt to a chatbot that returns five rows, two of them near duplicates, one quietly breaking your schema, and none of them the same when you ask again tomorrow. You came to build a feature, and instead you are maintaining fake data.
Here is the part most people miss: the expensive thing about test data is not the rows. It is deciding what a good row looks like. Do that thinking once, capture it, and you never pay for it again. That is the whole trick.
That is what a template is. You describe the data you want in a sentence, and you get a generator you can run forever:
pip install dugalaxy
