Sources

.hydra/sources/entries/ holds the material your data comes from: one record per source, usually an FtM Article with a title, publisher and link. Each file is a JSON list, like a collection:

[
  {
    "id": "wp-gus",
    "schema": "Article",
    "properties": {
      "title": [
        "Gus Fring"
      ],
      "publisher": [
        "Wikipedia"
      ],
      "sourceUrl": [
        "https://en.wikipedia.org/wiki/Gus_Fring"
      ]
    }
  }
]

A record cites its sources by id in its sources field. The link lives once, on the source entry, never repeated on the records.

Relationships (Ownership, Membership and the like) cite nothing; the entities they join carry the sources.

Two kinds of document

A source entry is where content came from: a Wikipedia page, a news article. It belongs in .hydra/sources/entries/.

A document in the data is part of the content itself: a case file and its pages that the investigation is about. It belongs in a collection, like any other record.

Neither appears as a box in the tree; both show up as evidence on the records that cite them.

Looking at sources

hydra-ctl sources draws what cites what, and the word sources after claims, node find or tree lists the sources behind just that result:

hydra-ctl sources "Walter Hartwell White"

sources

  entities   1

  documents

  HEISENBERG INVESTIGATION — DEA/ABQ CASE FILE (2008-2010) (Pages)   dea:doc-bb-casefile
  └── p.1 (Page)   heisenberg:doc-bb-casefile~p1
      └── Walter Hartwell White (Person)   sources   heisenberg:p-walt

  uncited    0

Sourcing one value at a time

A source in FtM is cited by a whole record, not by one of its fields, so Hydra can say which record cites a source but not which value the source supports.

To source values one by one, write one record per source with only what that source says. The build merges them, and claims then traces each value to its own record and source.