Pick a schema, toggle one or more of the 19 chaos operations, and see clean output and chaos-corrupted output rendered side‑by‑side with the changed fields highlighted.
Built for stress-testing the things downstream of fake data — ETL jobs, stream consumers, dashboards, alerting rules — under realistic failure conditions. Self-hosted via Docker Compose. Source on GitHub.
Pick a schema, select at least one chaos op, and click Generate. The same seeded run is executed twice — once with no chaos, once with the ops you selected — so non-chaos fields stay identical and only the corrupted fields differ.
Every interaction on this page is one HTTP call. The full deployment also ships WebSocket streaming, Redis + Postgres persistence, Kafka / Pub/Sub publishing, and Prometheus metrics.
$ curl "/api/generate?schema=smoke&count=10"
Three ops in this demo are full-deployment-only:
burst (WebSocket streaming),
schema_drift and data_drift (need Redis-backed state).