Skip to content

Drift monitoring

Ontoforge integrates with drift-monitor for production drift surveillance.

How it composes

production data  ──►  drift-monitor  ──►  ObservationRecord
                                            │
                                            ▼
                                       Ontoforge L10
                                            │
                                            ▼
                                       Evolution proposal
                                            │
                                            ▼
                                       Wizard step 9 (Evolution review)
  • drift-monitor computes PSI / KL / Wasserstein on each feature in your production pipeline.
  • It emits ObservationRecord JSON-LD objects whose IRI references the corresponding OWL class.
  • The Evolution phase reads those records and proposes ontology evolutions — new class, widened range, retired property.

Install the integration

Two steps — drift-monitor itself isn't on PyPI yet (it's installed straight from the synaptixs/infodrift repository), and PyPI policy bans pip install from pulling git+VCS URLs through a published package, so we keep the two halves separate:

pip install 'ontoforge[drift]'                                            # the PyPI half: drain3 + glue
pip install git+https://github.com/synaptixs/infodrift.git@develop       # drift-monitor itself

When drift-monitor ships to PyPI we'll fold this into a single install.

See examples/infodrift/ for a complete example.