ML & analytics training
Distribution-true samples drawn from pg_stats, the shape of production, without the data.
The problem
Models and dashboards need data that reflects real relationships and skew, but production data is blocked by privacy, access, or volume. Uniform random fixtures teach a model (or a chart) the wrong lessons.
How Weavori handles it
- Value distributions come from
pg_stats, so class balance and correlations carry over. - Referential integrity holds across joins, so feature tables stay consistent.
- Exports to CSV or JSON on stdout for notebooks and pipelines.
Export a training sample
Install the CLI
$npm install -g @weavori/cli
Generate and export
$weavori generate postgres://warehouse --rows 1000000 --output data.csv
Get started
i
Under the hood
The how it works guide explains distribution sampling and where it falls back to random.