Phase 12 — collocation coach: a third suggestion family for gentle
"natives usually say…" hints on non-native word pairings, reusing the
existing runPass/pendingScope/rail machinery.
- llm/collocation.go (RunCollocation, 25s floor, reuses ParseCheckpoint)
+ collocationSystemPrompt/CollocationMessages (warm, Mandarin gloss,
defers grammar/spelling to the grammar family)
- migration 0005 rebuilds the suggestions table to extend the type CHECK
(SQLite can't ALTER a CHECK)
- collocationScope + CollocationLimit + POST /{id}/collocation
- fix: grammarScope was `type != 'voice'` and would wipe the new
collocation flags; now `type NOT IN ('voice','collocation')`
- frontend: --color-blossom, "Make it sound natural 🌸" pill,
collocating/runCollocation in useCheckpoint, StatusBar dot
Phase 13 — vocabulary garden: capture looked-up words and surface them
for gentle spaced repetition.
- new internal/vocab package: migration 0006 (vocab_words, SM-2-lite
columns, doc_id ON DELETE SET NULL, UNIQUE(user_id,word)),
scheduler.go (Leitner ladder 1/3/7/16/35 then geometric; gentle
"again", no streak-shaming), handlers (capture-upsert/list/due/
review/delete, owner-scoped, SQLite-side datetime math)
- auto-capture on word lookup (dictionary-known words only, captures
the surrounding sentence + doc_id) + 🤍/💚 toggle on WordCard
- GardenPanel: blossom grid (bloom by reps), flashcard review (sentence
blanked, flip, again/good/easy, recognition↔production), sleepy-kitten
footer; opened from a global 🌷 header button
Tests: TestCollocationPassCoexists, vocab scheduler + handlers, db CHECK
extended. go build/vet/test + tsc + vite + vitest (51/51) clean;
migration verified against a copy of the live DB; live backend smoke
walked the full vocab lifecycle + the warm-502 collocation path.
Claude-Session: https://claude.ai/code/session_016Yr6jELuRc7hyzYLccQKZd
98 lines
3.0 KiB
Go
98 lines
3.0 KiB
Go
// Package vocab implements the vocabulary garden: it captures words the writer
|
|
// looks up and schedules them for gentle spaced-repetition review. The scheduler
|
|
// is a deliberately forgiving SM-2-lite / Leitner hybrid — no streaks to break,
|
|
// no harsh resets beyond a single step back — because this is a confidence
|
|
// builder, not a drill sergeant.
|
|
package vocab
|
|
|
|
import "math"
|
|
|
|
// Grade is the writer's self-assessment after seeing a flashcard's answer.
|
|
type Grade string
|
|
|
|
const (
|
|
GradeAgain Grade = "again" // didn't recall it — show again soon
|
|
GradeGood Grade = "good" // recalled it — advance one rung
|
|
GradeEasy Grade = "easy" // knew it instantly — advance a little further
|
|
)
|
|
|
|
// ladder is the Leitner interval ladder in days for the first several successful
|
|
// reviews: 1 → 3 → 7 → 16 → 35. Beyond the ladder, intervals grow by the card's
|
|
// ease factor so well-known words drift far into the future.
|
|
var ladder = []int{1, 3, 7, 16, 35}
|
|
|
|
const (
|
|
minEase = 1.3
|
|
startEase = 2.5
|
|
easyBonus = 1.3 // multiplier applied on top of a "good" step for "easy"
|
|
easeAgainDelta = -0.20 // ease nudged down on a lapse (still floored at minEase)
|
|
easeEasyDelta = 0.15 // ease nudged up when a card feels easy
|
|
)
|
|
|
|
// State is the spaced-repetition state of one card. It mirrors the scheduling
|
|
// columns on vocab_words so a review is "load State → next(grade) → persist".
|
|
type State struct {
|
|
Reps int
|
|
Interval int // days until the next review
|
|
Ease float64
|
|
Lapses int
|
|
}
|
|
|
|
// next returns the card's state after a review with the given grade. It never
|
|
// mutates the receiver. "again" steps the card back to a 1-day interval and
|
|
// counts a lapse (but only nudges ease down, never wipes progress harshly);
|
|
// "good" advances one rung of the ladder; "easy" advances a rung and a bit more.
|
|
func (s State) next(grade Grade) State {
|
|
ease := s.Ease
|
|
if ease == 0 {
|
|
ease = startEase
|
|
}
|
|
|
|
switch grade {
|
|
case GradeAgain:
|
|
return State{
|
|
Reps: 0,
|
|
Interval: 1,
|
|
Ease: clampEase(ease + easeAgainDelta),
|
|
Lapses: s.Lapses + 1,
|
|
}
|
|
case GradeEasy:
|
|
ease = clampEase(ease + easeEasyDelta)
|
|
reps := s.Reps + 1
|
|
return State{
|
|
Reps: reps,
|
|
Interval: int(math.Round(float64(goodInterval(reps, s.Interval, ease)) * easyBonus)),
|
|
Ease: ease,
|
|
Lapses: s.Lapses,
|
|
}
|
|
default: // GradeGood
|
|
reps := s.Reps + 1
|
|
return State{
|
|
Reps: reps,
|
|
Interval: goodInterval(reps, s.Interval, ease),
|
|
Ease: ease,
|
|
Lapses: s.Lapses,
|
|
}
|
|
}
|
|
}
|
|
|
|
// goodInterval returns the day-interval for a successful review: the explicit
|
|
// ladder while it lasts, then geometric growth by ease once past it.
|
|
func goodInterval(reps, prevInterval int, ease float64) int {
|
|
if reps >= 1 && reps <= len(ladder) {
|
|
return ladder[reps-1]
|
|
}
|
|
prev := prevInterval
|
|
if prev < ladder[len(ladder)-1] {
|
|
prev = ladder[len(ladder)-1]
|
|
}
|
|
return int(math.Round(float64(prev) * ease))
|
|
}
|
|
|
|
func clampEase(e float64) float64 {
|
|
if e < minEase {
|
|
return minEase
|
|
}
|
|
return e
|
|
}
|