Test Data Factory: synthetic fixtures for any schema

Three ways to generate

• From a preset topic: pick one of about thirty-five topics and choose the columns you want.
• From a schema: paste a CREATE TABLE statement, a CSV header row, a JSON sample, or a plain-English description, and the generator maps it to field types. An optional local model can help infer types; there is no external AI call.
• As a linked pair: generate a parent table and a child table joined by a real foreign key, so referential integrity holds in the output.

Deterministic and locale-real

Every run is seeded. The same specification and the same seed string reproduce byte-identical rows, so a fixture can be pinned across sprints and a failing test is reproducible. Data is locale-real across sixteen countries, up to five per run, with phone, currency, city, state and postal code that agree within a single row rather than a US ZIP stapled to an Indian address.

Identifiers that pass a validator but are not real

National and financial identifiers are emitted with valid checksums using the real algorithms a downstream validator checks, including Luhn, Verhoeff for Aadhaar, ISO 7064 mod-97 for IBAN, and the GSTIN and CPF check digits. Where a rule allows it, the value is made provably fictitious, for example an Aadhaar that begins with a digit the authority never issues. This is generated from rules and a seed, with no real personal data as input.

The quality report, export, templates and CI keys

• Each run returns a quality report computed by re-validating the emitted rows: uniqueness, per-identifier checksum validity, geo consistency, date rules, foreign-key integrity, distributions and null counts, plus plain statements you can quote in a review.
• Preview fifty rows, then download the full dataset regenerated server-side from the specification and seed, as JSON, CSV, Excel, SQL inserts (PostgreSQL, MySQL, SQLite or SQL Server dialects), or XML. Nothing the browser sends becomes the file.
• Save a specification as a re-runnable template, and mint a CI API key to run the same generation from a pipeline. Keys are shown once and draw on the same monthly allowance as the web app.

Limits worth knowing

• Rows per run scale by plan: Free 1,000, Solo Pro 10,000, Team and above 50,000. An over-cap request is clamped and you are told, not refused; every other capability is identical across plans.
• A run allows at most five countries, sixty custom fields and a fifty-thousand-row ceiling. A third simultaneous generation is asked to retry after a few seconds, because generation is memory-bounded on the shared host.
• Geography is cross-field verified for the United States and India; the other fourteen countries are locale-correct in format but not cross-checked, and the report labels which is which.