LOGOI

Documentation

Downloads & exports

The license, plainly

Everything the estate computed itself — frequency tables, attestation lists, co-occurrence tables, etymology pointers, receipts, coverage inventories — is yours to download and reuse under CC BY 4.0. The one condition is the attribution line every file carries. It names the exact receipt URL when that receipt has been materialized and says that the receipt is unavailable when it has not; a child result's receipt is never promoted to cover a composite comparison. Corpus text is different: a passage's Greek (or Latin, Hebrew, …) keeps the license of its source edition, stated per record, and is included only where that license permits public display.

What you can download

  • GET /api/v1/export/term/{language}/{lemma}.json — the per-word record. Morphology pages behind ?include=morphology&page=N so the default file stays small.
  • GET /api/v1/export/concordance/{language}/{lemma}.json|.csv — paginated attestations (?page=&page_size=&work_id=&work_ids=). Use either one exact canonical work_id, such as work:lat:ovid:phi006-perseus-lat2, or an ordered comma-separated cohort of 2–12 exact same-wing work_ids; never use a title or slug fragment.
  • GET /api/v1/export/frequency/{language}/{lemma}.json|.csv — per-work counts, optionally scoped by the same exact ?work_id=, normalized per 10,000 attested tokens, with a totals row in the CSV.
  • GET /api/v1/export/frequency-comparison/{language}/{left}/{right}.json|.csv — two complete frequency profiles joined over the same corpus wing and attested-token denominators, optionally scoped by exact ?work_id=. The comparison is descriptive: it does not claim statistical significance, lexical equivalence, or diachronic change.
  • GET /api/v1/export/cooccurrence/{language}/{lemma}.json|.csv — shared-passage co-occurrence (?lemma_b=) or companion ranking by distinct shared passages with complete passage marginals, Dice, and Jaccard. These descriptive set measures do not claim token proximity, order, significance, causation, equivalence, or diachronic change.
  • GET /api/v1/export/passage/{stable_id}.json — the passage record, with the text where its license permits.
  • GET /api/v1/export/source/{id}.json and receipt/{id}.json — provenance and audit records.
  • GET /api/v1/export/cite/{term|passage|source}/….bib — BibTeX; the JSON exports also carry CSL-JSON and a plain citation string.
  • GET /api/v1/export/tool/{tool}.json?… — any corpus tool result as a file, through the same interface agents use.
  • /downloads — per-work bulk JSONL of open-licensed text and estate-derived columns, with a manifest.

The machine-readable map is at /api/v1/export/index.json. Every tool payload from the MCP connector and the ask instruments now carries an export_url pointing at its own download.

Why some things stay excerpt-only

The licensed reference works — Beekes, Chantraine, Frisk, BDAG, HALOT, the OED lineage, Loeb apparatus — are cited, never reprinted. In every export path their material appears only as pointers and short excerpts with the authority, locator, and entry number, so a reader who holds the source can find the entry. This is enforced in the export layer itself, not by policy alone: citation-only prose is clamped to excerpt length even if a stored row were ever wider. Open lexica (LSJ under its open release) ship whole, with their own license stamped.

How to cite what you download

Each JSON export carries a cite block (BibTeX, CSL-JSON, and a plain string). For a passage, the citation names the stored edition — editor, publisher, year — with the Logoi receipt URL in the note. See the citation policy for the receipts law.