Lexicon Garden

at.kotoba.lesson.pack

Lesson content the app drills. Append-only stack; latest = active. Free-form by design: a known core the UI reads + open `meta`/`defaults` round-tripped untouched (SPEC §3.1a). This record body IS the upload interchange shape (lessons/pack.schema.json).

Repository @kotoba.at did:plc:o5f3cgna5sku4p6iitld6mxs namespace authority
Details · record, key tid
Kindrecord
Record keytid
Authoritylesson.kotoba.at
Indexed
Compare Edit in builder → Fork →

Record

record · key tid

Properties

createdAt

string

ISO 8601. NOTE: the app currently stamps a date-only string (YYYY-MM-DD), not a datetime — left unformatted here pending reconciliation (see README).

defaults

ref
Reference

items

requiredarray
itemsref
Reference

meta

unknown

Pack-level open extension; UI ignores, round-trips.

phase

string

Open. kana | vocab | grammar.

rationale

string

script

array

Declared scripts; drives chips + Focus. The scripts the glyphs are WRITTEN in — for a `word` item that is its kana script, not "word" (#99). Derived the same way if omitted; a word that is unsegmentable or mixes both kana falls back to "word". A declared script is cleaned by the app (#99): "word", junk and duplicates are dropped and the rest kept; "word" returns only if derivation finds an item with no derivable script.

itemsstring

title

requiredstring

#defaults

object

Pack-level drill defaults (SPEC §4). Open additional keys round-trip via the pack `meta`.

Properties

katakanaMultiplier

integer

Queue weight multiplier for katakana. Fixed-point ×10000 (e.g. 1.5 -> 15000); no float type.

tapToTypeBox

integer

Leitner box at/above which a glyph may graduate tap -> type.

  • minimum: 1
  • maximum: 5

#forms

object

Authored conjugation surfaces, keyed by the form names in src/drill/conjugate.ts. Kana only — the kanji surface is derived from the item's `kanji` form. A key this build does not know is ignored, so a form added later needs no lexicon change.

Properties

chau

string

imperative

string

masu

string

nai

string

nakatta

string

nakya

string

potential

string

ta

string

te

string

teru

string

toku

string

volitional

string

#item

object

A card. A discriminated union on `type` (#1), flattened into one object because lexicons have no oneOf: a RECOGNITION item (hiragana | katakana | kanji | word) requires glyph + romaji, a GRAMMAR item (particle | contraction | usage | sentence) requires id + prompt + answer. `conjugation` is a RECOGNITION type (#146): a derived verb surface, so it carries glyph + romaji like any word. Only `type` can be required here; the per-branch requiredness is enforced by lessons/pack.schema.json (JSON Schema if/then) and src/packs/validate.ts, which is what every reader of this record already routes through.

Properties

accept

array

GRAMMAR branch, optional. Alternative answers graded correct beside `answer` — casual forms routinely have two natural ones. Exact-match grading is unchanged; the accepted set just has more members.

itemsstring

answer

string

GRAMMAR branch, required there. The one canonical answer; graded by exact compare after trim.

choices

array

GRAMMAR branch, optional. Authored options for the choice grid (#145); contains `answer` when present. Never empty — the app drops an empty or wrong-typed list at the upload boundary rather than writing one.

itemsstring

confusableWith

array

ROMAJI (not glyphs) of same-script confusers; cold-start distractor seed.

itemsstring

forms

ref

Authored kana surfaces for the forms no rule derives (#136) — an irregular's whole paradigm, or one form of an otherwise regular verb (iku's itte). A form with no entry falls back to the derivation.

Reference

gloss

string

English meaning; recommended for words/kanji, and for a grammar item it is the meaning of the whole line.

glyph

string

RECOGNITION branch, required there. 1+ chars: single kana, word, or kanji.

group

integer

Verb conjugation class (#136): 1 godan, 2 ichidan, 3 irregular. Word items only. The app derives every form from it (src/drill/conjugate.ts) instead of the pack authoring them.

  • minimum: 1
  • maximum: 3

id

string

GRAMMAR branch, required there and unique within the pack. The SRS state map is glyph-keyed and a grammar item has no glyph, so it is keyed `grammar:<id>` (#145). Uniqueness is enforced by validate.ts, not here.

kanji

string

Written form in kanji (or kanji + okurigana) of a kana word — にほんじん -> 日本人. The below-hint path is passive (#125); a usable ruby meaning prompt records the separate kanji-recognition lane (#171). Omit where there is no kanji form (gairaigo) or the word is normally written in kana.

meta

unknown

Item-level open extension; UI ignores, round-trips.

prompt

string

GRAMMAR branch, required there. The sentence, with the blank written by the author (U+FF3F). Opaque to this layer: it may carry `{kanji|kana}` ruby markup (#139).

romaji

string

RECOGNITION branch, required there. Expected reading; lowercase Hepburn, exact case-insensitive match.

row

string

Gojuon row label, single kana and combination kana (yoon + extended) only.

ruby

string

Optional authored furigana boundaries for a recognition item, using `{base|reading}` markup. Used only when it reconstructs the plain `kanji` form and `glyph` reading; unusable markup falls back safely.

type

requiredstring

Open. Recognition: hiragana | katakana | kanji | word | conjugation (#146, a derived verb surface). Grammar (#1): particle | contraction | usage | sentence.

weight

integer

Optional author priority override. Fixed-point ×10000 (e.g. 1.8 -> 18000); lexicons have no float type. App scales at the record edge (see README).

Similar lexicons

semantic match

Lexicons whose schemas describe something close to this one. Check them before publishing a new one that overlaps.

  • at.kotoba.session

    record

    One practice session: the full per-card interaction array, the config that drove it, and session-level rollups. Append-only stack; TID rkeys (time-ordered).