BEQ filter-record contract¶
Version 1 defines one JSON object representing one published BEQ. A producer
writes one such object per file and a database.json array containing those
objects. These are source records, not the final public
docs/database.json: BEQCatalogue consumes source records, generates pages,
and adds its derived fields (catalogue_url, normalised audio fields and
filterAuthor) to the global database.
Required fields are title, year, audioTypes, content_type, author,
filters, mv, digest, created_at and updated_at. Producers must not
write catalogue_url; that belongs to BEQCatalogue's generated output.
content_type is film or TV. audioTypes is an array of strings.
filters is an array of miniDSP-style biquads. Each item has type, freq,
gain, q; shelf items also have count. Its biquads["96000"] object has
three b and two a coefficient strings. The array is unrolled: a repeated
filter appears once per biquad (and a shelf's count is therefore 1).
Optional metadata fields are altTitle, sortTitle, edition, season,
episode, note, warning, language, source, overview, rating,
runtime, collection, genres, avs, theMovieDB, images and
filterAuthor. Empty optional values may be omitted.
digest is SHA-256 over JSON containing exactly title, filters, mv,
season and episode, using the catalogue's canonical serialisation.
created_at is retained when a record is updated; updated_at changes when
its digest changes. Both are Unix seconds. The individual filename is not part
of the record contract.
An aggregate database.json is a JSON array of the individual records, sorted
by their individual relative path. It is derived output: consumers must treat
the individual records as authoritative, and producers must rewrite it whenever
they write, update or remove a record.