Source metadata records
Evidence, with the labels attached
Where the project
actually stands.
Source inventory phase · Snapshot checked September 30, 2026. These numbers describe the repository, not model performance.
Official opinion document links
Processed documents
Indexed retrieval chunks
| Layer | Status | What that means |
|---|---|---|
| Source inventory | In progress | 54 records: 45 discovery/catalog records and 9 official opinion links. |
| Text processing | Scaffold exists | No processed corpus yet. |
| Retrieval | Scaffold exists | Train-only sparse index builder; no indexed chunks yet. |
| Justice profiles | Provisional | Templates still require source grounding. |
| Prediction evaluation | Not run | No measured accuracy or calibration results. |
| Open-source release | Planned | Public repository and licensing details are not yet published here. |
A link is not a corpus.
A metadata record can point to an index, search page, or document. It does not establish that the text has been downloaded, cleaned, checked, or made available to a model. We report those stages separately so growth in an inventory cannot be mistaken for stronger predictions.
What we will measure
The evaluation plan includes vote agreement, citation faithfulness, retrieval quality, calibration, and temporal consistency. All are planned evaluation dimensions; none has a reported result yet.