Documentation

IntactSynth is on-premise software that produces a statistically faithful synthetic copy of a relational database. It reads your schema — tables, columns, foreign keys, check constraints, indexes, triggers — through read-only metadata queries, then generates new rows table by table in topological order. Referential integrity is validated at 100% before anything is exported; if validation fails, nothing is written.

It never reads production data values, makes no outbound network calls, and sends no telemetry. The output preserves the distributions, correlations and rare-event patterns of your real data while containing no information traceable to any real individual — which is what makes it usable for AI/ML training, software testing and onboarding without a KVKK or GDPR exposure.

A run is one command. The pipeline has four stages: (1) Schema Intelligence — connect with read-only metadata queries and map every object; (2) Semantic Analysis — infer what each column means and learn business rules from check constraints and trigger code; (3) Statistical Synthesis — generate rows table by table in dependency order, matching real-world distributions, breaking cycles with a deferred-constraint strategy inside a transaction; (4) Integrity Guarantee — the Quality Guardian validates every foreign key, every check constraint and every declared business rule before export.

IntactSynth ships on-premise only. It runs entirely inside your perimeter — no cloud APIs, no telemetry, no phone-home. License validation uses an offline, signed-token model: a license file scoped to your organization and server count is checked locally, with no activation server to reach.

The LLM backend is pluggable. An internet-connected deployment can use a hosted model; an air-gapped one runs a local model on your own GPU. Air-gapped deployment, on-site installation, the local LLM backend and source-code escrow are available on the Regulated tier.

Synthetic data produced by IntactSynth is not personal data — it contains no information traceable to any real individual — so it falls outside the definition of personal data under KVKK and the GDPR. GDPR Article 89 explicitly recognizes processing for statistical and research purposes. The generation process is documented so your compliance team can attach it to their records of processing and audit trails.

Relevant frameworks: KVKK (no personal data processed), GDPR Article 89 (statistical / research exemption), HIPAA Safe Harbor (de-identification), PCI-DSS (no cardholder data used). A formal document for each is available on request — see Resources.