Trey t dce2cc1f51 Make Full Table level-agnostic, fix the streak system end-to-end
Full Table (issue from chat): drop the level filter — Full Table tests
regular conjugation patterns, not vocabulary recognition, so restricting
to Basic-level verbs collapsed the eligible pool to two combos
(vivir present, ir future). Pool now draws from all 1,750 verbs. Random
sampling first; if 40 attempts fail we fall through to a deterministic
shuffled scan that guarantees finding any eligible (verb, tense) combo
when one exists. Returning nil now happens only when the user's filters
genuinely produce zero eligible prompts. The view replaces its silent
blank screen with a ContentUnavailableView pointing at the settings
that need adjusting. FeatureReferenceView documents the level exception.

Streak (issue #31 follow-up): activity recording was scoped to flashcard
and Full Table reviews only, so spending an hour on textbook work,
guides, videos, or AI chat could break a "streak" that the dashboard
kept displaying as if it were intact. Three fixes:

  1. Extract ReviewStore.recordActivity(context:) — a streak-only entry
     point that any user-initiated learning action can call.
  2. Add UserProgress.validateStreakIfStale(today:context:) — resets a
     broken currentStreak to 0 immediately, called from app launch and
     dashboard appear so the displayed number is never a lie.
  3. DailyLog formatter pins POSIX locale + current timezone so the
     yyyy-MM-dd strings can't drift across locales.

Wired recordActivity into every previously-silent learning action: chat
send, story-quiz completion, textbook exercise submit, grammar exercise
completion, course-deck study finish, week test / checkpoint save,
listening + pronunciation check, cloze quiz completion, lyrics word
lookup, video stream / play / download success, sentence-builder check,
and course-vocab SRS rate (which was bypassing ReviewStore entirely).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-26 01:24:27 -05:00
2026-04-21 23:31:14 -05:00
2026-04-23 07:33:41 -05:00

Conjuga

A Spanish-learning iOS app that combines verb conjugation practice, a full textbook reader, an AI conversation partner, offline dictionary lookups, grammar exercises, and more. Runs entirely on-device where possible (Foundation Models, Speech framework, Vision OCR).

Features

Verb practice

  • Six conjugation modes — flashcards, typing, multiple choice, handwriting (Apple Pencil / finger), sentence builder, and full-table (conjugate all persons at once)
  • Focus modes — weak verbs (SM-2 spaced repetition), irregularity drills (spelling / stem / unique), common tenses
  • 1,750 verbs across 5 levels (Basic → Expert) with 209 K pre-conjugated forms and 23 K irregular-span annotations
  • 20 tenses — every indicative, subjunctive, conditional, and imperative tense, each with character-level irregular highlighting
  • Verb list filters — search by level, by irregularity (Any / Spelling Change / Stem Change / Unique Irregular), or by reflexive only (curated 100-verb list); filters compose
  • Practice pool filters (Settings) — multi-select per level, per tense, per irregular-type, plus a "reflexive verbs only" toggle. Practice pool = intersection of all four.
  • Verb detail pages — conjugation table, six AI-generated example sentences (one per core tense), and reflexive infinitive + preposition hint for curated reflexive verbs
  • Text-to-speech on any form

Content & study

  • Textbook reader — 30 chapters of Complete Spanish Step-by-Step with 251 interactive exercises (keyboard + Apple Pencil), 931 OCR'd vocab tables rendered as Spanish→English grids (~3 100 paired cards extracted via bounding-box OCR)
  • Course decks — weekly vocab decks with example sentences, week tests, and cumulative checkpoint exams
  • Stem-change toggle on Week 4 flashcard decks (E-IE, E-I, O-UE, U-UE) showing inline present-tense conjugations
  • Grammar guide — 20 tense guides with usage rules and examples + 36 grammar topic notes (ser/estar, por/para, preterite/imperfect, reflexives, subjunctive triggers, etc.), each with 100+ practice exercises
  • Grammar exercises — interactive quizzes for 5 core topics (ser/estar, por/para, preterite/imperfect, subjunctive, personal a)
  • Curated YouTube videos — 54 hand-picked videos attached to guide and grammar items (preference for The Language Tutor's numbered lessons + BaseLang). Each item offers three actions: Stream (opens YouTube app / Safari), Download (YouTubeKit extraction + AVFoundation mux for modern adaptive-stream videos), Play (full-screen AVPlayer from local MP4). Settings → Downloaded Videos lists all downloads with total size, per-item delete, and a 500 MB warning.

AI & speech

  • Conversational practice — on-device AI chat partner (Apple Foundation Models) with 10 scenario types; chat bubbles have tappable words that open dictionary / on-demand AI lookup
  • AI short stories — generated stories with tappable words and comprehension quizzes
  • Listening practice — listen-and-type + pronunciation scoring via the Speech framework
  • Pronunciation check — word-by-word match scoring

Vocabulary & tools

  • Offline dictionary — reverse index of 175 K verb forms + 200 common words, cached to disk for instant lookups
  • Vocab SRS review — spaced repetition over course vocabulary with Again / Hard / Good / Easy rating
  • Cloze practice — fill-in-the-blank sentences with distractor generation
  • Lyrics practice — search, translate, and read Spanish song lyrics; long-press any word for an instant definition + tense/person readout (for verbs)

Tracking & sync

  • Progress — streaks, daily goals, accuracy stats, achievement badges, study-time tracking per day
  • CloudKit sync — review progress, test results, saved stories, conversations, and textbook attempts sync across devices
  • Widgets — combined dashboard + word-of-the-day, refreshed daily and on backgrounding

Architecture

  • SwiftUI + SwiftData with a dual-store configuration:
    • Local store (App Group group.com.conjuga.app) — reference data + per-device artifacts: verbs, forms, irregular spans, tense guides, course decks, vocab cards, textbook chapters, and downloaded videos. Seeded from bundled JSON on first launch. Self-healing re-seeds trigger on version bumps or if rows are missing on disk.
    • Cloud store (CloudKit iCloud.com.conjuga.app, private database) — user data: review cards, course reviews, user progress, test results, daily logs, saved songs, stories, conversations, textbook exercise attempts.
  • SharedModels Swift Package shared between the app and widget extension. Widget schema must include every local-store entity or SwiftData destructively migrates the shared store.
  • Foundation Models for on-device AI generation (@Generable structs for typed output) — conversation partner, short stories, verb-detail example sentences.
  • Vision framework for OCR of textbook pages and vocabulary images.
  • Speech framework for recognition and pronunciation scoring.
  • YouTubeKit (SPM) for video stream URL extraction, paired with AVFoundation (AVMutableComposition + AVAssetExportSession passthrough) to mux separate DASH video + audio tracks into a single MP4 when progressive streams aren't available.
  • Textbook extraction pipeline (Conjuga/Scripts/textbook/) — XHTML and answer-key parsers, macOS Vision image OCR + PDF page OCR, bounding-box vocab pair extractor, NSSpellChecker-based validator, and language-aware auto-fixer.
  • Video curation tooling (Conjuga/Scripts/generate_videos_markdown.py) — regenerates Conjuga/youtube_videos.md with per-video channel, upload date, duration, view count, and like count (pulled via yt-dlp).

Requirements

  • iOS 26+ (Foundation Models, Liquid Glass, modern Swift concurrency)
  • Xcode 26+
  • Apple Intelligence-capable device for AI features (conversation partner, AI stories, verb-detail examples); other features degrade gracefully

Building

Open Conjuga/Conjuga.xcodeproj in Xcode and run on a simulator or device. Reference data seeds automatically on first launch. To regenerate textbook content, run Conjuga/Scripts/textbook/run_pipeline.sh locally — the generated textbook_data.json / textbook_vocab.json are committed so fresh clones build without the pipeline.

S
Description
No description provided
Readme 24 MiB
Languages
Swift 86.1%
Python 11.9%
HTML 1.7%
Shell 0.3%