privaci 1.0.3

In-VPC PostgreSQL masking and anonymization engine

PrivaCI

One command. Sanitized staging data. No data leaves your VPC.

PrivaCI is a stateless batch engine that reads from a PostgreSQL source (typically a production replica), masks PII with a three-tier pipeline, and writes realistic synthetic data to a staging database with referential integrity preserved.

Prerequisites

• Python 3.12+, or the official container image (ghcr.io/boundarylogic/privaci)
• A PostgreSQL source (typically a production replica) and an empty target database

Quickstart

Fastest path — self-contained evaluation stack (~60 s):

export ANONYMIZATION_SALT="$(openssl rand -hex 32)"
make eval-up

See the quickstart for the full walkthrough.

Your own databases:

pip install -e .
privaci gen-salt > .privaci-salt && chmod 600 .privaci-salt
export ANONYMIZATION_SALT=$(cat .privaci-salt)
export SOURCE_DB_URL=postgresql://user:pass@source-host:5432/app
export TARGET_DB_URL=postgresql://user:pass@target-host:5432/staging

privaci validate && privaci dry-run && privaci run && privaci verify

Browse the docs site locally: pip install -e ".[dev]" && make docs-serve

Documentation

Start with the documentation site or quickstart. Key pages:

Using PrivaCI

• CLI reference — every command, its options, and examples
• Configuration reference — the mask-rules.yaml format
• Error codes — exit codes and message format
• State & audit schema — what runs write to _privaci
• Extending PrivaCI — the plugin contract model

Developing PrivaCI

• Local development & testing
• Test fixtures — MedicalHelpDesk Corp
• Architecture decision records

License

The engine is licensed under the Elastic License 2.0. Optional paid features ship as a separate plugin layer via the plugin contracts.

Commercial

Optional paid capabilities — signed compliance reports, schema-drift detection, FK-aware subsetting, JSONB path masking, and more — ship as a separate plugin layer on top of this engine, still entirely in your VPC. Learn more at BoundaryLogic.