Open source
Apache-2.0, developed in the open, and built for Build with DataHub: The Agent Hackathon.
Contributing
The ground rules
Never fake an integration
A fixture may stand in for DataHub in a test. It must never be presented as a live connection at runtime.
The engine stays deterministic
A model may help classify intent. It must never be the thing that decides allow, ask, or deny.
Fail safe, not open
Any new code path that can fail needs a documented decision for the failure case, and a test asserting it.
Extractors need false-positive tests
An extractor that over-reports causes approval fatigue, which is its own failure mode. Precision is a safety property here.
Locally
Working on it
git clone https://github.com/AmirmLotfy/zence && cd zence
uv sync --all-packages
uv run pytest -m "not integration and not e2e"
uv run ruff check . && uv run mypyDisclosure
How this was built
Zence was developed with AI assistance (Claude Code) during the hackathon submission period. No pre-existing project code was incorporated; third-party frameworks and libraries are declared in pyproject.toml and package.json. The fictional clients, datasets and people in the demo catalog are synthetic.