Leya API Reference

Returns {query, results[], result_count, limit, truncated}. Results carry document_id, source_id, title, snippet (~500 chars), score, url, effective_from, effective_to, publication_date and source_last_indexed.

Pagination. When has_more is true, pass next_cursor back as cursor with q and every filter unchanged. The cursor is a keyset position in one query's score ordering, so it carries a fingerprint of the query that produced it — replaying it elsewhere is a 422, not a plausible wrong page. Only claim an exhaustive search once has_more is false.

Empty results mean no coverage. country, document_type and legal_area are validated against the corpus: an unknown value is a 422 listing what's accepted, so a zero-result page is genuine absence rather than a typo. source_id is not validated this way — take those from /v1/sources.

Full text of one document. Pass a document_id from a search result verbatim. Returns body_markdown plus has_body and body_length.

has_body: false means the document is indexed but its text has not been extracted. That is not an empty document, and it is not a reason to supply the text from memory — report it as a gap. Bodies are unbounded; check body_length before loading a full code into a constrained context.

A citation string for inline use, plus the components it was built from.

{
  "formatted": "Decreto 1441, Congreso de la República (1961)",
  "style": "latam-standard",
  "document_type": "decreto",
  "number": "1441",
  "issuing_body": "Congreso de la República",
  "year": 1961
}

style: latam-standard is the full <Tipo> <Número>, <Órgano Emisor> (<Año>) form. style: title-fallback means the corpus lacks that metadata, so the string is title plus id and the citation is partial.

Use formatted verbatim in both cases. Completing a partial citation from memory produces a fabricated authority — the specific failure this API exists to prevent. A null component means unknown.

The values country, document_type and legal_area actually accept, each with a document count. Read this before filtering — it's the only way to tell a typo'd filter from real no-coverage.

{
  "countries":      [{"value": "GT",      "document_count": 960}],
  "document_types": [{"value": "decreto", "document_count": 412}],
  "legal_areas":    [{"value": "laboral", "document_count": 87}]
}

Registered datasets and the source_id values the search filter accepts.

Key introspection: the identity a key belongs to, its key_type, its rate_limits, and requests_last_30d. Read your budget from here rather than hard-coding the numbers above.

No auth. Per-source and per-country document counts, date spans, last ingest time and status. This is the authoritative answer to “what does Leya cover?” — prefer it over any static list. Rendered for humans at /status.

One cadastral parcel from Guatemala’s RIC, by código catastral (e.g. 18-04-09-03971). Returns the parcel, a coverage block, and overlaps[] — the protected areas, sacred sites, archaeological sites, watersheds and forestry-incentive zones it intersects, each with an overlap_fraction so containment is distinguishable from a boundary graze.

This is not a property registry. The ccc is RIC’s identifier, not the Registro General de la Propiedad’s finca/folio/libro, and no mapping between them is published. There is no owner, title or encumbrance data here.

RIC has surveyed a minority of the country — nine departments publish zero parcels — so a 404 usually means unsurveyed, not nonexistent. The error carries a coverage object saying which.

Per-department parcel counts: what the cadastral dataset does and does not contain. Read this before drawing any conclusion from an empty parcel result. See also /v1/parcels for listing and /v1/cadastre/layers for the overlay keys.