Add dictionary_entries seed tool + share JMdict field extraction #8

Merged

raisondetredev merged 1 commit from refs/pull/8/head into main

2026-06-05 20:26:47 +02:00

raisondetredev commented

2026-06-05 20:26:37 +02:00

(Migrated from codeberg.org)

DictionarySeedRow.fromWord builds a catalog row, reusing the same
expression/reading/meanings extraction as the card curator — now pulled
out into vocabulary_extraction.dart so the two can't drift (an 'Add to
my deck' card must match the curated deck's card for the same word).
search_blob holds every surface form, lowercased, for trigram lookup;
English glosses are excluded by design. Adds http; 11 new tests green,
existing curator suite unaffected by the extraction refactor.

Signed-off-by: Tizian 「ティツィアーン」 raisondetredev@pm.me

A new bin (halkyon_curator:seed_dictionary) loads a jmdict-simplified JSON file and upserts it into the server-side dictionary_entries catalog via PostgREST, batched, on the JMdict id primary key (merge-duplicates), so re-runs are idempotent. The service-role key is read from SUPABASE_SERVICE_ROLE_KEY (never argv); --dry-run builds and samples rows without sending. Optional --jlpt-dir sets each row's level; --common-only / --limit scope the seed. DictionarySeedRow.fromWord builds a catalog row, reusing the same expression/reading/meanings extraction as the card curator — now pulled out into vocabulary_extraction.dart so the two can't drift (an 'Add to my deck' card must match the curated deck's card for the same word). search_blob holds every surface form, lowercased, for trigram lookup; English glosses are excluded by design. Adds http; 11 new tests green, existing curator suite unaffected by the extraction refactor. Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>