Pipeline: adapters and ingest

What the pipeline does

We don’t invent a corpus. We build the machinery that lets anyone take their own jurisdiction’s already-published legal data and turn it into a common, verifiable, interoperable form, then keep it that way as the law changes.

The pipeline is that machinery. Three stages, one per jurisdiction.

1. Ingest

An adapter reads the official source: a national gazette feed, an open-data portal, a bulk export, whatever that country actually publishes. Adapters are thin on purpose. Each one knows the quirks of exactly one source and nothing else.

2. Validate

The adapter emits Akoma Ntoso that conforms to AKN4OLF, and the pipeline runs it against the conformance suite. Output that fails doesn’t get published. There is no “mostly correct” tier. That single rule is what keeps adapters written by people who have never met interoperable.

3. Publish

What passes flows into the archive, addressed by its stable OLF identifier and carrying its normalized temporal metadata and citation graph.

Provenance, not scraping

Every document the pipeline emits records where it came from and how it was built. The goal isn’t to own a copy of the law. It’s to produce a copy whose lineage back to the official source is explicit, and reproducible: re-run the pipeline, get the same bytes. That is the whole difference between a scraped dump and a provenance layer.

Adding a jurisdiction

An adapter is a self-contained piece of work. Read one official source, emit AKN4OLF, pass the suite. You don’t coordinate with other jurisdictions and you don’t ask permission. The suite is the gatekeeper, not a committee.

Maintain legal data for a jurisdiction and want it here? Open an issue in pipeline.

Status

Early. The first two adapters are Italy and France: two civil-law systems, two very different gazettes. We picked them to find out whether one model really holds across more than one legal tradition before we claim it’s general.