Supporting app · beta

DataMaker

Spec-driven synthetic data — deterministic where it should be, LLM only for language.

Read the guide GitHub

DataMaker (also shipped as DataScribe) interviews you into a generation spec, then seeds tables with reproducible generators. Numbers, ids, and time series stay in code; only semantic text hits a model.

Why it sits next to Zero

Crews are only as good as the context they can read. DataMaker produces JSON datasets you can drop into a Zero workspace so specialists cite realistic tables instead of inventing numbers.

Typical loop

  1. Generate a dataset in DataMaker (export JSON)
  2. Upload into a Zero organization (or sync via your pipeline)
  3. Ask the crew questions that require those tables
  4. Optionally inspect the same DB/file with pgLens

See Bring your data into Zero.