Fixes #28 — Curated reflexive verb list on detail + practice filter

Bundles the 100 most common reflexive verbs from spanishwithdaniel.com as a
canonical list and wires it through the UI. Compound list entries (recibirse
/ graduarse, equivocarse / confundirse) are split. Trailing prepositions and
set-phrase completions are captured as usageHint (e.g. acordarse "de",
ponerse "de acuerdo").

ReflexiveVerbStore loads the JSON at launch and exposes lookups by base
infinitive, both via @Environment for SwiftUI and a static shared instance
for services. Verbs whose bare infinitive isn't in the list skip the UI
treatment silently.

VerbDetailView shows a new Reflexive section with the reflexive infinitive,
usage hint, and English meaning when there is a match. VerbListView gains a
"Reflexive verbs only" filter alongside the existing Level and Irregularity
filters. Settings adds the same flag so it also constrains the practice
pool; PracticeSessionService applies the reflexive filter in all six pick
paths.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

.gitignore

@@ -40,3 +40,16 @@ scrape/
 *.webm
 *.mp4
 *.mkv
+
+# Third-party textbook sources (not redistributable)
+*.pdf
+*.epub
+epub_extract/
+
+# Textbook extraction artifacts — regenerate locally via run_pipeline.sh.
+# Scripts are committed; their generated outputs are not.
+Conjuga/Scripts/textbook/*.json
+Conjuga/Scripts/textbook/review.html
+# Note: the app-bundle copies (Conjuga/Conjuga/textbook_{data,vocab}.json)
+# ARE committed so `xcodebuild` works on a fresh clone without first running
+# the pipeline. They're regenerated from the scripts when content changes.

CLAUDE.md

@@ -0,0 +1,5 @@
+# Project rules
+
+## Git
+
+- **Never run `git commit` or `git push` without an explicit request from the user in the current turn.** File edits are fine; committing and pushing are not. Wait to be told.

Conjuga/Conjuga.xcodeproj/project.pbxproj

@@ -8,10 +8,15 @@
 
 /* Begin PBXBuildFile section */
 		00BEC0BDBB49198022D9852E /* WordOfDayWidget.swift in Sources */ = {isa = PBXBuildFile; fileRef = 8E9BCDBB9BC24F5C8117767E /* WordOfDayWidget.swift */; };
+		04C74D9E0ED84BF785A8331C /* ClozeView.swift in Sources */ = {isa = PBXBuildFile; fileRef = D232CDA43CC9218D748BA121 /* ClozeView.swift */; };
 		0A89DCC82BE11605CB866DEF /* TenseInfo.swift in Sources */ = {isa = PBXBuildFile; fileRef = 3BC3247457109FC6BF00D85B /* TenseInfo.swift */; };
+		0AD63CAED7C568590A16E879 /* StoryQuizView.swift in Sources */ = {isa = PBXBuildFile; fileRef = A7CDC5F2660A3009A3ADF048 /* StoryQuizView.swift */; };
+		0D0D3B5CC128D1A1D1252282 /* PronunciationService.swift in Sources */ = {isa = PBXBuildFile; fileRef = 4AE3D1D8723D5C41D3501774 /* PronunciationService.swift */; };
 		13F29AD5745FB532709FA28A /* OnboardingView.swift in Sources */ = {isa = PBXBuildFile; fileRef = E972AA745F44586EF0B1B0C8 /* OnboardingView.swift */; };
+		14242FD1F500D296D41E927C /* FeatureReferenceView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 1C3E36BDC2540AF2A67AEEB1 /* FeatureReferenceView.swift */; };
 		1A230C01A045F0C095BFBD35 /* PracticeView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 1EA0FA4F9149B9D8E197ADE9 /* PracticeView.swift */; };
 		1C2636790E70B6BC7FFCC904 /* DailyLog.swift in Sources */ = {isa = PBXBuildFile; fileRef = 0313D24F96E6A0039C34341F /* DailyLog.swift */; };
+		20B71403A8D305C29C73ADA2 /* StemChangeConjugationView.swift in Sources */ = {isa = PBXBuildFile; fileRef = F92BCE1A6720E47FCD26BADC /* StemChangeConjugationView.swift */; };
 		218E982FC4267949F82AABAD /* SharedModels in Frameworks */ = {isa = PBXBuildFile; productRef = 4A4D7B02884EBA9ACD93F0FD /* SharedModels */; };
 		261E582449BED6EF41881B04 /* AdaptiveContainer.swift in Sources */ = {isa = PBXBuildFile; fileRef = 3B16FF4C52457CD8CD703532 /* AdaptiveContainer.swift */; };
 		27BA7FA9356467846A07697D /* TypingView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 10C16AA6022E4742898745CE /* TypingView.swift */; };
@@ -20,6 +25,7 @@
 		2C7ABAB4D88E3E3B0EAD1EF7 /* PracticeHeaderView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5BF946245110C92F087D81E8 /* PracticeHeaderView.swift */; };
 		33E885EB38C3BB0CB058871A /* HandwritingView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 1F842EB5E566C74658D918BB /* HandwritingView.swift */; };
 		352A5BAA6E406AA5850653A4 /* PracticeSessionService.swift in Sources */ = {isa = PBXBuildFile; fileRef = 842DB48F8570C39CDCFF2F57 /* PracticeSessionService.swift */; };
+		354631F309E625046A3A436B /* TextbookExerciseView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 854EA2A8D6CF203958BA3C24 /* TextbookExerciseView.swift */; };
 		35A0F6E7124D989312721F7D /* DashboardView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 18AC3C548BDB9EF8701BE64C /* DashboardView.swift */; };
 		36F92EBAEB0E5F2B010401EF /* StreakCalendarView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 30EF2362D9FFF9B07A45CE6D /* StreakCalendarView.swift */; };
 		377C4AA000CE9A0D8CC43DA9 /* GrammarNote.swift in Sources */ = {isa = PBXBuildFile; fileRef = 4D389CA5B5C4E7A12CAEA5BC /* GrammarNote.swift */; };
@@ -27,32 +33,50 @@
 		3F4F0C07BE61512CBFBBB203 /* HandwritingCanvas.swift in Sources */ = {isa = PBXBuildFile; fileRef = 80D974250C396589656B8443 /* HandwritingCanvas.swift */; };
 		4005E258FDF03C8B3A0D53BD /* VocabFlashcardView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 2931634BEB33B93429CE254F /* VocabFlashcardView.swift */; };
 		46943ACFABF329DE1CBFC471 /* TensePill.swift in Sources */ = {isa = PBXBuildFile; fileRef = 102F0E136CDFF8CED710210F /* TensePill.swift */; };
+		48967E05C65E32F7082716CD /* AnswerChecker.swift in Sources */ = {isa = PBXBuildFile; fileRef = B3EFFA19D0AB2528A868E8ED /* AnswerChecker.swift */; };
 		4C3484403FD96E37DA4BEA66 /* NewWordIntent.swift in Sources */ = {isa = PBXBuildFile; fileRef = 72CB5F95DF256DF7CD73269D /* NewWordIntent.swift */; };
+		4C577CF6B137D0A32759A169 /* VerbExampleGenerator.swift in Sources */ = {isa = PBXBuildFile; fileRef = 02EB3F9305349775E0EB28B9 /* VerbExampleGenerator.swift */; };
 		50E0095A23E119D1AB561232 /* VerbDetailView.swift in Sources */ = {isa = PBXBuildFile; fileRef = E1DBE662F89F02A0282F5BEE /* VerbDetailView.swift */; };
 		519E68D2DF4C80AB96058C0D /* LyricsConfirmationView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 3EA01795655C444795577A22 /* LyricsConfirmationView.swift */; };
 		51D072AF30F4B12CD3E8F918 /* SRSEngine.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5C0E6EAFC0D24928BA956FA5 /* SRSEngine.swift */; };
 		53A0AC57EAC44B676C997374 /* QuizType.swift in Sources */ = {isa = PBXBuildFile; fileRef = 626873572466403C0288090D /* QuizType.swift */; };
 		5A3246026E68AB6483126D0B /* WeekProgressWidget.swift in Sources */ = {isa = PBXBuildFile; fileRef = 1980E8E439EB76ED7330A90D /* WeekProgressWidget.swift */; };
 		5EA915FFA906C5C2938FCADA /* ConjugaWidgetBundle.swift in Sources */ = {isa = PBXBuildFile; fileRef = E325FE0E484DE75009672D02 /* ConjugaWidgetBundle.swift */; };
+		5EE41911F3D17224CAB359ED /* StudyTimerService.swift in Sources */ = {isa = PBXBuildFile; fileRef = 4EC8C4E931AD7A1D87C490BB /* StudyTimerService.swift */; };
 		60E86BABE2735E2052B99DF3 /* SettingsView.swift in Sources */ = {isa = PBXBuildFile; fileRef = BCCC95A95581458E068E0484 /* SettingsView.swift */; };
+		61328552866DE185B15011A9 /* StoryLibraryView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 15AC27B1E3D332709657F20B /* StoryLibraryView.swift */; };
 		615D3128ED6E84EF59BB5AA3 /* LyricsReaderView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 58394296923991E56BAC2B02 /* LyricsReaderView.swift */; };
 		6BB4B0A655E6CB6F82D81B5A /* WeekTestView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5E7EF4161C73AAC67B3A0004 /* WeekTestView.swift */; };
-		968D626462B0ADEC8D7D56AA /* CheckpointExamView.swift in Sources */ = {isa = PBXBuildFile; fileRef = EA1F177F7ABF5D2E4E5466CD /* CheckpointExamView.swift */; };
 		6D4A29280FDD99B8E18AF264 /* WidgetDataReader.swift in Sources */ = {isa = PBXBuildFile; fileRef = 2889F2F81673AFF3A58A07A8 /* WidgetDataReader.swift */; };
 		6ED2AC2CAA54688161D4B920 /* SyncStatusMonitor.swift in Sources */ = {isa = PBXBuildFile; fileRef = 18CCD69C14D1B0CFBD03C92F /* SyncStatusMonitor.swift */; };
 		728702D9AA7A8BDABBA62513 /* ReviewStore.swift in Sources */ = {isa = PBXBuildFile; fileRef = CBCF6FCFA6B00151C2371E77 /* ReviewStore.swift */; };
 		760628EFE1CF191CE2FC07DC /* GuideView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 8C935ECDF8A5D8D6FA541E20 /* GuideView.swift */; };
+		78FE99C5D511737B6877EDD5 /* VocabReviewView.swift in Sources */ = {isa = PBXBuildFile; fileRef = E8D95887B18216FCA71643D6 /* VocabReviewView.swift */; };
 		7A13757EA40E81E55640D0FC /* LyricsSearchView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 70960F0FD7509310B3F61C48 /* LyricsSearchView.swift */; };
+		81E4DB9F64F3FF3AB8BCB03A /* TextbookChapterListView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 496D851D2D95BEA283C9FD45 /* TextbookChapterListView.swift */; };
 		81FA7EBCF18F0AAE0BF385C3 /* VerbListView.swift in Sources */ = {isa = PBXBuildFile; fileRef = A63061BBC8998DF33E3DCA2B /* VerbListView.swift */; };
 		82F6079BE3F31AC3FB2D1013 /* MultipleChoiceView.swift in Sources */ = {isa = PBXBuildFile; fileRef = DA3A33983B2F2078C9EA1A3D /* MultipleChoiceView.swift */; };
 		84CCBAE22A9E0DA27AE28723 /* DeckStudyView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 631DC0A942DD57C81DECE083 /* DeckStudyView.swift */; };
+		8B516215E0842189DEA0DBB1 /* GrammarExercise.swift in Sources */ = {isa = PBXBuildFile; fileRef = F0A3099BE24A56F9B1F179E0 /* GrammarExercise.swift */; };
+		8BD4B5A2DDDD4BE4B4A94962 /* ChatView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 79576893566932D2BE207528 /* ChatView.swift */; };
 		8C43F09F52EA9B537EA27E43 /* CourseReviewStore.swift in Sources */ = {isa = PBXBuildFile; fileRef = DAF7CA1E6F9979CB2C699FDC /* CourseReviewStore.swift */; };
+		8DC1CB93333F94C5297D33BF /* GrammarExerciseView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 07F24ED76059609D6857EC97 /* GrammarExerciseView.swift */; };
+		90B76E34F195223580F7CCCF /* DictionaryService.swift in Sources */ = {isa = PBXBuildFile; fileRef = 76BE2A08EC694FF784ED5575 /* DictionaryService.swift */; };
 		943A94A8C71919F3EFC0E8FA /* UserProgress.swift in Sources */ = {isa = PBXBuildFile; fileRef = E536AD1180FE10576EAC884A /* UserProgress.swift */; };
+		97A2088134FC6CB41C507182 /* reflexive_verbs.json in Resources */ = {isa = PBXBuildFile; fileRef = 3644B5ED77F29A65877D926A /* reflexive_verbs.json */; };
 		97EFCF6724CE59DC4F0274FD /* AchievementService.swift in Sources */ = {isa = PBXBuildFile; fileRef = 1C42EA0EBD4CB1E10A82BA25 /* AchievementService.swift */; };
 		9D9FD3853C5C969C62AE9999 /* StartupCoordinator.swift in Sources */ = {isa = PBXBuildFile; fileRef = A4B95B276C054DBFE508C4D1 /* StartupCoordinator.swift */; };
+		9F0ACDC1F4ACB1E0D331283D /* CheckpointExamView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 34C67DD1A1CB9B8B5A2BDCED /* CheckpointExamView.swift */; };
+		A4DA8CF1957A4FE161830AB2 /* ReflexiveVerbStore.swift in Sources */ = {isa = PBXBuildFile; fileRef = 940826D9ED5C18D2C4E7B2C7 /* ReflexiveVerbStore.swift */; };
+		A7DF435F99E66E067F2B33E1 /* ListeningView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 20D1904DF07E0A6816134CF3 /* ListeningView.swift */; };
 		A9959AE6C87B4AD21554E401 /* FullTableView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 711CB7539EF5887F6F7B8B82 /* FullTableView.swift */; };
 		AAC6F85A1C3B6C1186E1656A /* TenseEndingTable.swift in Sources */ = {isa = PBXBuildFile; fileRef = 69D98E1564C6538056D81200 /* TenseEndingTable.swift */; };
+		ABBE5080E254D1D3E3465E40 /* ConversationService.swift in Sources */ = {isa = PBXBuildFile; fileRef = 3A96C065B8787DEC6818E497 /* ConversationService.swift */; };
+		ACE9D8B3116757B5D6F0F766 /* StoryGenerator.swift in Sources */ = {isa = PBXBuildFile; fileRef = 713F23A9C2935408B136C7C7 /* StoryGenerator.swift */; };
+		B10A324C06F0957DDE2233F8 /* TextbookChapterView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 39908548430FDF01D76201FB /* TextbookChapterView.swift */; };
 		B4603AA6EFB134794AA39BF4 /* LyricsLibraryView.swift in Sources */ = {isa = PBXBuildFile; fileRef = FC2B1F646394D7C03493F1BF /* LyricsLibraryView.swift */; };
+		B48C0015BE53279B0631C2D7 /* ChatLibraryView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 648436F8326CF95777E2FA58 /* ChatLibraryView.swift */; };
+		BA3DE2DA319AA3B572C19E11 /* VerbExampleCache.swift in Sources */ = {isa = PBXBuildFile; fileRef = EBEEC9CC9A8C502AF5F42914 /* VerbExampleCache.swift */; };
 		BB48230C3B26EA6E84D2D823 /* DailyProgressRing.swift in Sources */ = {isa = PBXBuildFile; fileRef = 180F9D59828C36B44A5E384F /* DailyProgressRing.swift */; };
 		BF0832865857EFDA1D1CDEAD /* SharedModels in Frameworks */ = {isa = PBXBuildFile; productRef = BCCBABD74CADDB118179D8E9 /* SharedModels */; };
 		C0BAEF49A6270D8F64CF13D6 /* PracticeViewModel.swift in Sources */ = {isa = PBXBuildFile; fileRef = C359C051FB157EF447561405 /* PracticeViewModel.swift */; };
@@ -61,6 +85,7 @@
 		C3851F960C1162239DC2F935 /* CourseQuizView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 143D06606AE10DCA30A140C2 /* CourseQuizView.swift */; };
 		C8C3880535008764B7117049 /* DataLoader.swift in Sources */ = {isa = PBXBuildFile; fileRef = DADCA82DDD34DF36D59BB283 /* DataLoader.swift */; };
 		CAC69045B74249F121643E88 /* AnswerReviewView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 83A8C1A048627C8DEB83C12D /* AnswerReviewView.swift */; };
+		CC886125F8ECE72D1AAD4861 /* StoryReaderView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 6A48474D969CEF5F573DF09B /* StoryReaderView.swift */; };
 		CF9E48ADF0501FB79F3DDB7B /* conjuga_data.json in Resources */ = {isa = PBXBuildFile; fileRef = 8C2D88FF9A3B0590B22C7837 /* conjuga_data.json */; };
 		D3FFE73A5AD27F1759F50727 /* SpeechService.swift in Sources */ = {isa = PBXBuildFile; fileRef = 49E3AD244327CBF24B7A2752 /* SpeechService.swift */; };
 		D40B4E919DE379C50265CA9F /* SyncToast.swift in Sources */ = {isa = PBXBuildFile; fileRef = 1C4B5204F6B8647C816814F0 /* SyncToast.swift */; };
@@ -78,22 +103,6 @@
 		F59655A8B8FCE6264315DD33 /* Assets.xcassets in Resources */ = {isa = PBXBuildFile; fileRef = A014EEC3EE08E945FBBA5335 /* Assets.xcassets */; };
 		F84706B47A2156B2138FB8D5 /* GrammarNotesView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 3F1A6221A35699BD8065D064 /* GrammarNotesView.swift */; };
 		FC7873F97017532C215DAD34 /* ReviewCard.swift in Sources */ = {isa = PBXBuildFile; fileRef = 0A8A63F750065CA4EF36B4D3 /* ReviewCard.swift */; };
-		DDF58F3899FC4B92BF6587D2 /* StudyTimerService.swift in Sources */ = {isa = PBXBuildFile; fileRef = 978FB24DF8D7436CB5210ACE /* StudyTimerService.swift */; };
-		8C1E4E7F36D64EFF8D092AC8 /* StoryGenerator.swift in Sources */ = {isa = PBXBuildFile; fileRef = 327659ABFD524514B6D2D505 /* StoryGenerator.swift */; };
-		4C2649215B81470195F38ED0 /* StoryLibraryView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 950347251CC94D4A9DFF7CBC /* StoryLibraryView.swift */; };
-		8E3D8E8254CF4213B9D9FAD3 /* StoryReaderView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 2A8B6081226847E0A0A174BC /* StoryReaderView.swift */; };
-		12D2C9311D5C4764B48B1754 /* StoryQuizView.swift in Sources */ = {isa = PBXBuildFile; fileRef = E292A183ABB24FFE9CB719C8 /* StoryQuizView.swift */; };
-		8D7CA0F4496B44C28CD5EBD5 /* DictionaryService.swift in Sources */ = {isa = PBXBuildFile; fileRef = A04370CF6B4E4D38BE3EB0C7 /* DictionaryService.swift */; };
-		3EC2A2F4B9C24B029DA49C40 /* VocabReviewView.swift in Sources */ = {isa = PBXBuildFile; fileRef = D3698CE7ACF148318615293E /* VocabReviewView.swift */; };
-		53908E41767B438C8BD229CD /* ClozeView.swift in Sources */ = {isa = PBXBuildFile; fileRef = A649B04B8B3C49419AD9219C /* ClozeView.swift */; };
-		65ABC39F35804C619DAB3200 /* GrammarExercise.swift in Sources */ = {isa = PBXBuildFile; fileRef = 17E5252282F44ECD9BA70DB8 /* GrammarExercise.swift */; };
-		B73F6EED00304B718C6FEFFA /* GrammarExerciseView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 1F71CA5CD67342F18319DB9A /* GrammarExerciseView.swift */; };
-		EA07DB964C8940F69C14DE2C /* PronunciationService.swift in Sources */ = {isa = PBXBuildFile; fileRef = D535EF6988A24B47B70209A2 /* PronunciationService.swift */; };
-		4DCC5CC233DE4701A12FD7EB /* ListeningView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 02B2179562E54E148C98219D /* ListeningView.swift */; };
-		35D6404C60C249D5995AD895 /* ConversationService.swift in Sources */ = {isa = PBXBuildFile; fileRef = E10603F454E54341AA4B9931 /* ConversationService.swift */; };
-		C8AF0931F7FD458C80B6EC0D /* ChatLibraryView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 5667AA04211A449A9150BD28 /* ChatLibraryView.swift */; };
-		6CCC8D51F5524688A4BC5AF8 /* ChatView.swift in Sources */ = {isa = PBXBuildFile; fileRef = FA5FE6E149F54A6BA7D01D99 /* ChatView.swift */; };
-		8510085D78E248D885181E80 /* FeatureReferenceView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 12E9DDEFD53C49E0A48EA655 /* FeatureReferenceView.swift */; };
 /* End PBXBuildFile section */
 
 /* Begin PBXContainerItemProxy section */
@@ -121,26 +130,35 @@
 /* End PBXCopyFilesBuildPhase section */
 
 /* Begin PBXFileReference section */
+		02EB3F9305349775E0EB28B9 /* VerbExampleGenerator.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = VerbExampleGenerator.swift; sourceTree = "<group>"; };
 		0313D24F96E6A0039C34341F /* DailyLog.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = DailyLog.swift; sourceTree = "<group>"; };
+		07F24ED76059609D6857EC97 /* GrammarExerciseView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = GrammarExerciseView.swift; sourceTree = "<group>"; };
 		0A8A63F750065CA4EF36B4D3 /* ReviewCard.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ReviewCard.swift; sourceTree = "<group>"; };
 		102F0E136CDFF8CED710210F /* TensePill.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = TensePill.swift; sourceTree = "<group>"; };
 		10C16AA6022E4742898745CE /* TypingView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = TypingView.swift; sourceTree = "<group>"; };
 		143D06606AE10DCA30A140C2 /* CourseQuizView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = CourseQuizView.swift; sourceTree = "<group>"; };
+		15AC27B1E3D332709657F20B /* StoryLibraryView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = StoryLibraryView.swift; sourceTree = "<group>"; };
 		16C1F74196C3C5628953BE3F /* Conjuga.app */ = {isa = PBXFileReference; includeInIndex = 0; lastKnownFileType = wrapper.application; path = Conjuga.app; sourceTree = BUILT_PRODUCTS_DIR; };
 		180F9D59828C36B44A5E384F /* DailyProgressRing.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = DailyProgressRing.swift; sourceTree = "<group>"; };
 		18AC3C548BDB9EF8701BE64C /* DashboardView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = DashboardView.swift; sourceTree = "<group>"; };
 		18CCD69C14D1B0CFBD03C92F /* SyncStatusMonitor.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = SyncStatusMonitor.swift; sourceTree = "<group>"; };
 		195DA9CDA703DDFAD1B3CD5A /* DailyProgressWidget.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = DailyProgressWidget.swift; sourceTree = "<group>"; };
 		1980E8E439EB76ED7330A90D /* WeekProgressWidget.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = WeekProgressWidget.swift; sourceTree = "<group>"; };
+		1C3E36BDC2540AF2A67AEEB1 /* FeatureReferenceView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = FeatureReferenceView.swift; sourceTree = "<group>"; };
 		1C42EA0EBD4CB1E10A82BA25 /* AchievementService.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = AchievementService.swift; sourceTree = "<group>"; };
 		1C4B5204F6B8647C816814F0 /* SyncToast.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = SyncToast.swift; sourceTree = "<group>"; };
 		1EA0FA4F9149B9D8E197ADE9 /* PracticeView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = PracticeView.swift; sourceTree = "<group>"; };
 		1EB4830F9289AACC82D753F8 /* ConjugaApp.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ConjugaApp.swift; sourceTree = "<group>"; };
 		1F842EB5E566C74658D918BB /* HandwritingView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = HandwritingView.swift; sourceTree = "<group>"; };
+		20D1904DF07E0A6816134CF3 /* ListeningView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ListeningView.swift; sourceTree = "<group>"; };
 		2889F2F81673AFF3A58A07A8 /* WidgetDataReader.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = WidgetDataReader.swift; sourceTree = "<group>"; };
 		2931634BEB33B93429CE254F /* VocabFlashcardView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = VocabFlashcardView.swift; sourceTree = "<group>"; };
 		30EF2362D9FFF9B07A45CE6D /* StreakCalendarView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = StreakCalendarView.swift; sourceTree = "<group>"; };
+		34C67DD1A1CB9B8B5A2BDCED /* CheckpointExamView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = CheckpointExamView.swift; sourceTree = "<group>"; };
+		3644B5ED77F29A65877D926A /* reflexive_verbs.json */ = {isa = PBXFileReference; lastKnownFileType = text.json; path = reflexive_verbs.json; sourceTree = "<group>"; };
 		3695075616689E72DBB26D4C /* HandwritingRecognizer.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = HandwritingRecognizer.swift; sourceTree = "<group>"; };
+		39908548430FDF01D76201FB /* TextbookChapterView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = TextbookChapterView.swift; sourceTree = "<group>"; };
+		3A96C065B8787DEC6818E497 /* ConversationService.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ConversationService.swift; sourceTree = "<group>"; };
 		3B16FF4C52457CD8CD703532 /* AdaptiveContainer.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = AdaptiveContainer.swift; sourceTree = "<group>"; };
 		3BC3247457109FC6BF00D85B /* TenseInfo.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = TenseInfo.swift; sourceTree = "<group>"; };
 		3CC1AD23158CBABBB753FA1E /* ConjugaWidget.entitlements */ = {isa = PBXFileReference; lastKnownFileType = text.plist.entitlements; path = ConjugaWidget.entitlements; sourceTree = "<group>"; };
@@ -149,42 +167,54 @@
 		42ADC600530309A9B147A663 /* IrregularHighlightText.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = IrregularHighlightText.swift; sourceTree = "<group>"; };
 		43345D6C7EAA4017E3A45935 /* CombinedWidget.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = CombinedWidget.swift; sourceTree = "<group>"; };
 		43B8AED76C14A05AF2339C27 /* LyricsSearchService.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = LyricsSearchService.swift; sourceTree = "<group>"; };
+		496D851D2D95BEA283C9FD45 /* TextbookChapterListView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = TextbookChapterListView.swift; sourceTree = "<group>"; };
 		49E3AD244327CBF24B7A2752 /* SpeechService.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = SpeechService.swift; sourceTree = "<group>"; };
+		4AE3D1D8723D5C41D3501774 /* PronunciationService.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = PronunciationService.swift; sourceTree = "<group>"; };
 		4D389CA5B5C4E7A12CAEA5BC /* GrammarNote.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = GrammarNote.swift; sourceTree = "<group>"; };
+		4EC8C4E931AD7A1D87C490BB /* StudyTimerService.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = StudyTimerService.swift; sourceTree = "<group>"; };
 		58394296923991E56BAC2B02 /* LyricsReaderView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = LyricsReaderView.swift; sourceTree = "<group>"; };
 		5983A534E4836F30B5281ACB /* MainTabView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = MainTabView.swift; sourceTree = "<group>"; };
 		5BF946245110C92F087D81E8 /* PracticeHeaderView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = PracticeHeaderView.swift; sourceTree = "<group>"; };
 		5C0E6EAFC0D24928BA956FA5 /* SRSEngine.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = SRSEngine.swift; sourceTree = "<group>"; };
 		5E7EF4161C73AAC67B3A0004 /* WeekTestView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = WeekTestView.swift; sourceTree = "<group>"; };
-		EA1F177F7ABF5D2E4E5466CD /* CheckpointExamView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = CheckpointExamView.swift; sourceTree = "<group>"; };
 		626873572466403C0288090D /* QuizType.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = QuizType.swift; sourceTree = "<group>"; };
 		631DC0A942DD57C81DECE083 /* DeckStudyView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = DeckStudyView.swift; sourceTree = "<group>"; };
+		648436F8326CF95777E2FA58 /* ChatLibraryView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ChatLibraryView.swift; sourceTree = "<group>"; };
 		69D98E1564C6538056D81200 /* TenseEndingTable.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = TenseEndingTable.swift; sourceTree = "<group>"; };
+		6A48474D969CEF5F573DF09B /* StoryReaderView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = StoryReaderView.swift; sourceTree = "<group>"; };
 		6B9A9F2AB21895E06989A4D5 /* FlashcardView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = FlashcardView.swift; sourceTree = "<group>"; };
 		70960F0FD7509310B3F61C48 /* LyricsSearchView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = LyricsSearchView.swift; sourceTree = "<group>"; };
 		711CB7539EF5887F6F7B8B82 /* FullTableView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = FullTableView.swift; sourceTree = "<group>"; };
+		713F23A9C2935408B136C7C7 /* StoryGenerator.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = StoryGenerator.swift; sourceTree = "<group>"; };
 		72CB5F95DF256DF7CD73269D /* NewWordIntent.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = NewWordIntent.swift; sourceTree = "<group>"; };
 		731614CACCB73B6FD592D34A /* SentenceBuilderView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = SentenceBuilderView.swift; sourceTree = "<group>"; };
+		76BE2A08EC694FF784ED5575 /* DictionaryService.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = DictionaryService.swift; sourceTree = "<group>"; };
 		777C696A841803D5B775B678 /* ReferenceStore.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ReferenceStore.swift; sourceTree = "<group>"; };
+		79576893566932D2BE207528 /* ChatView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ChatView.swift; sourceTree = "<group>"; };
 		7E6AF62A3A949630E067DC22 /* Info.plist */ = {isa = PBXFileReference; lastKnownFileType = text.plist; path = Info.plist; sourceTree = "<group>"; };
 		80D974250C396589656B8443 /* HandwritingCanvas.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = HandwritingCanvas.swift; sourceTree = "<group>"; };
 		833516C5D57F164C8660A479 /* CourseView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = CourseView.swift; sourceTree = "<group>"; };
 		83A8C1A048627C8DEB83C12D /* AnswerReviewView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = AnswerReviewView.swift; sourceTree = "<group>"; };
 		842DB48F8570C39CDCFF2F57 /* PracticeSessionService.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = PracticeSessionService.swift; sourceTree = "<group>"; };
+		854EA2A8D6CF203958BA3C24 /* TextbookExerciseView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = TextbookExerciseView.swift; sourceTree = "<group>"; };
 		8C2D88FF9A3B0590B22C7837 /* conjuga_data.json */ = {isa = PBXFileReference; lastKnownFileType = text.json; path = conjuga_data.json; sourceTree = "<group>"; };
 		8C935ECDF8A5D8D6FA541E20 /* GuideView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = GuideView.swift; sourceTree = "<group>"; };
 		8E9BCDBB9BC24F5C8117767E /* WordOfDayWidget.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = WordOfDayWidget.swift; sourceTree = "<group>"; };
+		940826D9ED5C18D2C4E7B2C7 /* ReflexiveVerbStore.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ReflexiveVerbStore.swift; sourceTree = "<group>"; };
 		9708FF3CF33E4765DB225F93 /* ConjugaWidgetExtension.appex */ = {isa = PBXFileReference; includeInIndex = 0; lastKnownFileType = "wrapper.app-extension"; path = ConjugaWidgetExtension.appex; sourceTree = BUILT_PRODUCTS_DIR; };
 		9E1FB35614B709E6B1D1D017 /* Conjuga.entitlements */ = {isa = PBXFileReference; lastKnownFileType = text.plist.entitlements; path = Conjuga.entitlements; sourceTree = "<group>"; };
 		A014EEC3EE08E945FBBA5335 /* Assets.xcassets */ = {isa = PBXFileReference; lastKnownFileType = folder.assetcatalog; path = Assets.xcassets; sourceTree = "<group>"; };
 		A4B95B276C054DBFE508C4D1 /* StartupCoordinator.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = StartupCoordinator.swift; sourceTree = "<group>"; };
 		A63061BBC8998DF33E3DCA2B /* VerbListView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = VerbListView.swift; sourceTree = "<group>"; };
+		A7CDC5F2660A3009A3ADF048 /* StoryQuizView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = StoryQuizView.swift; sourceTree = "<group>"; };
 		AC34396050805693AA4AC582 /* Info.plist */ = {isa = PBXFileReference; lastKnownFileType = text.plist; path = Info.plist; sourceTree = "<group>"; };
+		B3EFFA19D0AB2528A868E8ED /* AnswerChecker.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = AnswerChecker.swift; sourceTree = "<group>"; };
 		BC273716CD14A99EFF8206CA /* course_data.json */ = {isa = PBXFileReference; lastKnownFileType = text.json; path = course_data.json; sourceTree = "<group>"; };
 		BCCC95A95581458E068E0484 /* SettingsView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = SettingsView.swift; sourceTree = "<group>"; };
 		C359C051FB157EF447561405 /* PracticeViewModel.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = PracticeViewModel.swift; sourceTree = "<group>"; };
 		CBCF6FCFA6B00151C2371E77 /* ReviewStore.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ReviewStore.swift; sourceTree = "<group>"; };
 		CF6D58AEE2F0DFE0F1829A73 /* SharedModels */ = {isa = PBXFileReference; lastKnownFileType = folder; name = SharedModels; path = SharedModels; sourceTree = SOURCE_ROOT; };
+		D232CDA43CC9218D748BA121 /* ClozeView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ClozeView.swift; sourceTree = "<group>"; };
 		D570252DA3DCDD9217C71863 /* WidgetDataService.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = WidgetDataService.swift; sourceTree = "<group>"; };
 		DA3A33983B2F2078C9EA1A3D /* MultipleChoiceView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = MultipleChoiceView.swift; sourceTree = "<group>"; };
 		DADCA82DDD34DF36D59BB283 /* DataLoader.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = DataLoader.swift; sourceTree = "<group>"; };
@@ -193,25 +223,13 @@
 		E1DBE662F89F02A0282F5BEE /* VerbDetailView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = VerbDetailView.swift; sourceTree = "<group>"; };
 		E325FE0E484DE75009672D02 /* ConjugaWidgetBundle.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ConjugaWidgetBundle.swift; sourceTree = "<group>"; };
 		E536AD1180FE10576EAC884A /* UserProgress.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = UserProgress.swift; sourceTree = "<group>"; };
+		E8D95887B18216FCA71643D6 /* VocabReviewView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = VocabReviewView.swift; sourceTree = "<group>"; };
 		E8E9833868EB73AF9EB3A611 /* StoreInspector.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = StoreInspector.swift; sourceTree = "<group>"; };
 		E972AA745F44586EF0B1B0C8 /* OnboardingView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = OnboardingView.swift; sourceTree = "<group>"; };
+		EBEEC9CC9A8C502AF5F42914 /* VerbExampleCache.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = VerbExampleCache.swift; sourceTree = "<group>"; };
+		F0A3099BE24A56F9B1F179E0 /* GrammarExercise.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = GrammarExercise.swift; sourceTree = "<group>"; };
+		F92BCE1A6720E47FCD26BADC /* StemChangeConjugationView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = StemChangeConjugationView.swift; sourceTree = "<group>"; };
 		FC2B1F646394D7C03493F1BF /* LyricsLibraryView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = LyricsLibraryView.swift; sourceTree = "<group>"; };
-		978FB24DF8D7436CB5210ACE /* StudyTimerService.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = StudyTimerService.swift; sourceTree = "<group>"; };
-		327659ABFD524514B6D2D505 /* StoryGenerator.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = StoryGenerator.swift; sourceTree = "<group>"; };
-		950347251CC94D4A9DFF7CBC /* StoryLibraryView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = StoryLibraryView.swift; sourceTree = "<group>"; };
-		2A8B6081226847E0A0A174BC /* StoryReaderView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = StoryReaderView.swift; sourceTree = "<group>"; };
-		E292A183ABB24FFE9CB719C8 /* StoryQuizView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = StoryQuizView.swift; sourceTree = "<group>"; };
-		A04370CF6B4E4D38BE3EB0C7 /* DictionaryService.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = DictionaryService.swift; sourceTree = "<group>"; };
-		D3698CE7ACF148318615293E /* VocabReviewView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = VocabReviewView.swift; sourceTree = "<group>"; };
-		A649B04B8B3C49419AD9219C /* ClozeView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ClozeView.swift; sourceTree = "<group>"; };
-		17E5252282F44ECD9BA70DB8 /* GrammarExercise.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = GrammarExercise.swift; sourceTree = "<group>"; };
-		1F71CA5CD67342F18319DB9A /* GrammarExerciseView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = GrammarExerciseView.swift; sourceTree = "<group>"; };
-		D535EF6988A24B47B70209A2 /* PronunciationService.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = PronunciationService.swift; sourceTree = "<group>"; };
-		02B2179562E54E148C98219D /* ListeningView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ListeningView.swift; sourceTree = "<group>"; };
-		E10603F454E54341AA4B9931 /* ConversationService.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ConversationService.swift; sourceTree = "<group>"; };
-		5667AA04211A449A9150BD28 /* ChatLibraryView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ChatLibraryView.swift; sourceTree = "<group>"; };
-		FA5FE6E149F54A6BA7D01D99 /* ChatView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ChatView.swift; sourceTree = "<group>"; };
-		12E9DDEFD53C49E0A48EA655 /* FeatureReferenceView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = FeatureReferenceView.swift; sourceTree = "<group>"; };
 /* End PBXFileReference section */
 
 /* Begin PBXFrameworksBuildPhase section */
@@ -243,6 +261,7 @@
 				1EB4830F9289AACC82D753F8 /* ConjugaApp.swift */,
 				BC273716CD14A99EFF8206CA /* course_data.json */,
 				7E6AF62A3A949630E067DC22 /* Info.plist */,
+				3644B5ED77F29A65877D926A /* reflexive_verbs.json */,
 				353C5DE41FD410FA82E3AED7 /* Models */,
 				1994867BC8E985795A172854 /* Services */,
 				BFC1AEBE02CE22E6474FFEA6 /* Utilities */,
@@ -255,8 +274,8 @@
 		0931AEB5B728C3A03F06A1CA /* Settings */ = {
 			isa = PBXGroup;
 			children = (
+				1C3E36BDC2540AF2A67AEEB1 /* FeatureReferenceView.swift */,
 				BCCC95A95581458E068E0484 /* SettingsView.swift */,
-				12E9DDEFD53C49E0A48EA655 /* FeatureReferenceView.swift */,
 			);
 			path = Settings;
 			sourceTree = "<group>";
@@ -274,23 +293,27 @@
 			isa = PBXGroup;
 			children = (
 				1C42EA0EBD4CB1E10A82BA25 /* AchievementService.swift */,
+				B3EFFA19D0AB2528A868E8ED /* AnswerChecker.swift */,
+				3A96C065B8787DEC6818E497 /* ConversationService.swift */,
 				DAF7CA1E6F9979CB2C699FDC /* CourseReviewStore.swift */,
 				DADCA82DDD34DF36D59BB283 /* DataLoader.swift */,
+				76BE2A08EC694FF784ED5575 /* DictionaryService.swift */,
 				3695075616689E72DBB26D4C /* HandwritingRecognizer.swift */,
 				43B8AED76C14A05AF2339C27 /* LyricsSearchService.swift */,
 				842DB48F8570C39CDCFF2F57 /* PracticeSessionService.swift */,
-				E10603F454E54341AA4B9931 /* ConversationService.swift */,
+				4AE3D1D8723D5C41D3501774 /* PronunciationService.swift */,
-				D535EF6988A24B47B70209A2 /* PronunciationService.swift */,
-				A04370CF6B4E4D38BE3EB0C7 /* DictionaryService.swift */,
-				327659ABFD524514B6D2D505 /* StoryGenerator.swift */,
-				978FB24DF8D7436CB5210ACE /* StudyTimerService.swift */,
 				777C696A841803D5B775B678 /* ReferenceStore.swift */,
+				940826D9ED5C18D2C4E7B2C7 /* ReflexiveVerbStore.swift */,
 				CBCF6FCFA6B00151C2371E77 /* ReviewStore.swift */,
 				49E3AD244327CBF24B7A2752 /* SpeechService.swift */,
 				5C0E6EAFC0D24928BA956FA5 /* SRSEngine.swift */,
 				A4B95B276C054DBFE508C4D1 /* StartupCoordinator.swift */,
 				E8E9833868EB73AF9EB3A611 /* StoreInspector.swift */,
+				713F23A9C2935408B136C7C7 /* StoryGenerator.swift */,
+				4EC8C4E931AD7A1D87C490BB /* StudyTimerService.swift */,
 				18CCD69C14D1B0CFBD03C92F /* SyncStatusMonitor.swift */,
+				EBEEC9CC9A8C502AF5F42914 /* VerbExampleCache.swift */,
+				02EB3F9305349775E0EB28B9 /* VerbExampleGenerator.swift */,
 				D570252DA3DCDD9217C71863 /* WidgetDataService.swift */,
 			);
 			path = Services;
@@ -308,6 +331,7 @@
 			isa = PBXGroup;
 			children = (
 				0313D24F96E6A0039C34341F /* DailyLog.swift */,
+				F0A3099BE24A56F9B1F179E0 /* GrammarExercise.swift */,
 				4D389CA5B5C4E7A12CAEA5BC /* GrammarNote.swift */,
 				626873572466403C0288090D /* QuizType.swift */,
 				0A8A63F750065CA4EF36B4D3 /* ReviewCard.swift */,
@@ -315,8 +339,7 @@
 				3BC3247457109FC6BF00D85B /* TenseInfo.swift */,
 				DAFE27F29412021AEC57E728 /* TestResult.swift */,
 				E536AD1180FE10576EAC884A /* UserProgress.swift */,
-							17E5252282F44ECD9BA70DB8 /* GrammarExercise.swift */,
+			);
-);
 			path = Models;
 			sourceTree = "<group>";
 		};
@@ -341,6 +364,16 @@
 			path = ViewModels;
 			sourceTree = "<group>";
 		};
+		43E4D263B0AF47E401A51601 /* Stories */ = {
+			isa = PBXGroup;
+			children = (
+				15AC27B1E3D332709657F20B /* StoryLibraryView.swift */,
+				A7CDC5F2660A3009A3ADF048 /* StoryQuizView.swift */,
+				6A48474D969CEF5F573DF09B /* StoryReaderView.swift */,
+			);
+			path = Stories;
+			sourceTree = "<group>";
+		};
 		4B183AB0C56BC2EC302531E7 /* ConjugaWidget */ = {
 			isa = PBXGroup;
 			children = (
@@ -361,52 +394,33 @@
 			isa = PBXGroup;
 			children = (
 				83A8C1A048627C8DEB83C12D /* AnswerReviewView.swift */,
+				D232CDA43CC9218D748BA121 /* ClozeView.swift */,
 				6B9A9F2AB21895E06989A4D5 /* FlashcardView.swift */,
 				711CB7539EF5887F6F7B8B82 /* FullTableView.swift */,
 				1F842EB5E566C74658D918BB /* HandwritingView.swift */,
+				20D1904DF07E0A6816134CF3 /* ListeningView.swift */,
 				DA3A33983B2F2078C9EA1A3D /* MultipleChoiceView.swift */,
 				5BF946245110C92F087D81E8 /* PracticeHeaderView.swift */,
 				1EA0FA4F9149B9D8E197ADE9 /* PracticeView.swift */,
-				D3698CE7ACF148318615293E /* VocabReviewView.swift */,
 				731614CACCB73B6FD592D34A /* SentenceBuilderView.swift */,
 				10C16AA6022E4742898745CE /* TypingView.swift */,
+				E8D95887B18216FCA71643D6 /* VocabReviewView.swift */,
+				8FB89F19B33894DDF27C8EC2 /* Chat */,
 				895E547BEFB5D0FBF676BE33 /* Lyrics */,
-				8A1DED0596E04DDE9536A9A9 /* Stories */,
+				43E4D263B0AF47E401A51601 /* Stories */,
-				DFD75E32A53845A693D98F48 /* Chat */,
+			);
-				02B2179562E54E148C98219D /* ListeningView.swift */,
-				A649B04B8B3C49419AD9219C /* ClozeView.swift */,
-);
 			path = Practice;
 			sourceTree = "<group>";
 		};
 		8102F7FA5BFE6D38B2212AD3 /* Guide */ = {
 			isa = PBXGroup;
 			children = (
+				07F24ED76059609D6857EC97 /* GrammarExerciseView.swift */,
 				3F1A6221A35699BD8065D064 /* GrammarNotesView.swift */,
 				8C935ECDF8A5D8D6FA541E20 /* GuideView.swift */,
-							1F71CA5CD67342F18319DB9A /* GrammarExerciseView.swift */,
+			);
-);
 			path = Guide;
 			sourceTree = "<group>";
-		};
-				DFD75E32A53845A693D98F48 /* Chat */ = {
-			isa = PBXGroup;
-			children = (
-				5667AA04211A449A9150BD28 /* ChatLibraryView.swift */,
-				FA5FE6E149F54A6BA7D01D99 /* ChatView.swift */,
-			);
-			path = Chat;
-			sourceTree = "<group>";
-		};
-		8A1DED0596E04DDE9536A9A9 /* Stories */ = {
-			isa = PBXGroup;
-			children = (
-				950347251CC94D4A9DFF7CBC /* StoryLibraryView.swift */,
-				2A8B6081226847E0A0A174BC /* StoryReaderView.swift */,
-				E292A183ABB24FFE9CB719C8 /* StoryQuizView.swift */,
-			);
-			path = Stories;
-			sourceTree = "<group>";
 		};
 		895E547BEFB5D0FBF676BE33 /* Lyrics */ = {
 			isa = PBXGroup;
@@ -419,6 +433,15 @@
 			path = Lyrics;
 			sourceTree = "<group>";
 		};
+		8FB89F19B33894DDF27C8EC2 /* Chat */ = {
+			isa = PBXGroup;
+			children = (
+				648436F8326CF95777E2FA58 /* ChatLibraryView.swift */,
+				79576893566932D2BE207528 /* ChatView.swift */,
+			);
+			path = Chat;
+			sourceTree = "<group>";
+		};
 		A591A3B6F1F13D23D68D7A9D = {
 			isa = PBXGroup;
 			children = (
@@ -457,12 +480,16 @@
 		BE5A40BAC9DD6884C58A2096 /* Course */ = {
 			isa = PBXGroup;
 			children = (
+				34C67DD1A1CB9B8B5A2BDCED /* CheckpointExamView.swift */,
 				143D06606AE10DCA30A140C2 /* CourseQuizView.swift */,
 				833516C5D57F164C8660A479 /* CourseView.swift */,
 				631DC0A942DD57C81DECE083 /* DeckStudyView.swift */,
+				F92BCE1A6720E47FCD26BADC /* StemChangeConjugationView.swift */,
+				496D851D2D95BEA283C9FD45 /* TextbookChapterListView.swift */,
+				39908548430FDF01D76201FB /* TextbookChapterView.swift */,
+				854EA2A8D6CF203958BA3C24 /* TextbookExerciseView.swift */,
 				2931634BEB33B93429CE254F /* VocabFlashcardView.swift */,
 				5E7EF4161C73AAC67B3A0004 /* WeekTestView.swift */,
-				EA1F177F7ABF5D2E4E5466CD /* CheckpointExamView.swift */,
 			);
 			path = Course;
 			sourceTree = "<group>";
@@ -585,6 +612,7 @@
 				F59655A8B8FCE6264315DD33 /* Assets.xcassets in Resources */,
 				CF9E48ADF0501FB79F3DDB7B /* conjuga_data.json in Resources */,
 				2B5B2D63DC9C290F66890A4A /* course_data.json in Resources */,
+				97A2088134FC6CB41C507182 /* reflexive_verbs.json in Resources */,
 			);
 			runOnlyForDeploymentPostprocessing = 0;
 		};
@@ -597,8 +625,14 @@
 			files = (
 				97EFCF6724CE59DC4F0274FD /* AchievementService.swift in Sources */,
 				261E582449BED6EF41881B04 /* AdaptiveContainer.swift in Sources */,
+				48967E05C65E32F7082716CD /* AnswerChecker.swift in Sources */,
 				CAC69045B74249F121643E88 /* AnswerReviewView.swift in Sources */,
+				B48C0015BE53279B0631C2D7 /* ChatLibraryView.swift in Sources */,
+				8BD4B5A2DDDD4BE4B4A94962 /* ChatView.swift in Sources */,
+				9F0ACDC1F4ACB1E0D331283D /* CheckpointExamView.swift in Sources */,
+				04C74D9E0ED84BF785A8331C /* ClozeView.swift in Sources */,
 				C2B3D97F119EFCE97E3CB1CE /* ConjugaApp.swift in Sources */,
+				ABBE5080E254D1D3E3465E40 /* ConversationService.swift in Sources */,
 				C3851F960C1162239DC2F935 /* CourseQuizView.swift in Sources */,
 				8C43F09F52EA9B537EA27E43 /* CourseReviewStore.swift in Sources */,
 				F0D0778207F144D6AC3D39C3 /* CourseView.swift in Sources */,
@@ -607,8 +641,12 @@
 				35A0F6E7124D989312721F7D /* DashboardView.swift in Sources */,
 				C8C3880535008764B7117049 /* DataLoader.swift in Sources */,
 				84CCBAE22A9E0DA27AE28723 /* DeckStudyView.swift in Sources */,
+				90B76E34F195223580F7CCCF /* DictionaryService.swift in Sources */,
+				14242FD1F500D296D41E927C /* FeatureReferenceView.swift in Sources */,
 				D4DDE25FB2DAD315370AFB74 /* FlashcardView.swift in Sources */,
 				A9959AE6C87B4AD21554E401 /* FullTableView.swift in Sources */,
+				8B516215E0842189DEA0DBB1 /* GrammarExercise.swift in Sources */,
+				8DC1CB93333F94C5297D33BF /* GrammarExerciseView.swift in Sources */,
 				377C4AA000CE9A0D8CC43DA9 /* GrammarNote.swift in Sources */,
 				F84706B47A2156B2138FB8D5 /* GrammarNotesView.swift in Sources */,
 				760628EFE1CF191CE2FC07DC /* GuideView.swift in Sources */,
@@ -616,6 +654,7 @@
 				E7BFEE9A90E1300EFF5B1F32 /* HandwritingRecognizer.swift in Sources */,
 				33E885EB38C3BB0CB058871A /* HandwritingView.swift in Sources */,
 				28D2F489F1927BCCC2B56086 /* IrregularHighlightText.swift in Sources */,
+				A7DF435F99E66E067F2B33E1 /* ListeningView.swift in Sources */,
 				519E68D2DF4C80AB96058C0D /* LyricsConfirmationView.swift in Sources */,
 				B4603AA6EFB134794AA39BF4 /* LyricsLibraryView.swift in Sources */,
 				615D3128ED6E84EF59BB5AA3 /* LyricsReaderView.swift in Sources */,
@@ -628,8 +667,10 @@
 				352A5BAA6E406AA5850653A4 /* PracticeSessionService.swift in Sources */,
 				1A230C01A045F0C095BFBD35 /* PracticeView.swift in Sources */,
 				C0BAEF49A6270D8F64CF13D6 /* PracticeViewModel.swift in Sources */,
+				0D0D3B5CC128D1A1D1252282 /* PronunciationService.swift in Sources */,
 				53A0AC57EAC44B676C997374 /* QuizType.swift in Sources */,
 				DF82C2579F9889DDB06362CC /* ReferenceStore.swift in Sources */,
+				A4DA8CF1957A4FE161830AB2 /* ReflexiveVerbStore.swift in Sources */,
 				FC7873F97017532C215DAD34 /* ReviewCard.swift in Sources */,
 				728702D9AA7A8BDABBA62513 /* ReviewStore.swift in Sources */,
 				51D072AF30F4B12CD3E8F918 /* SRSEngine.swift in Sources */,
@@ -637,39 +678,34 @@
 				60E86BABE2735E2052B99DF3 /* SettingsView.swift in Sources */,
 				D3FFE73A5AD27F1759F50727 /* SpeechService.swift in Sources */,
 				9D9FD3853C5C969C62AE9999 /* StartupCoordinator.swift in Sources */,
+				20B71403A8D305C29C73ADA2 /* StemChangeConjugationView.swift in Sources */,
 				E814A9CF1067313F74B509C6 /* StoreInspector.swift in Sources */,
+				ACE9D8B3116757B5D6F0F766 /* StoryGenerator.swift in Sources */,
+				61328552866DE185B15011A9 /* StoryLibraryView.swift in Sources */,
+				0AD63CAED7C568590A16E879 /* StoryQuizView.swift in Sources */,
+				CC886125F8ECE72D1AAD4861 /* StoryReaderView.swift in Sources */,
 				36F92EBAEB0E5F2B010401EF /* StreakCalendarView.swift in Sources */,
+				5EE41911F3D17224CAB359ED /* StudyTimerService.swift in Sources */,
 				6ED2AC2CAA54688161D4B920 /* SyncStatusMonitor.swift in Sources */,
 				D40B4E919DE379C50265CA9F /* SyncToast.swift in Sources */,
 				AAC6F85A1C3B6C1186E1656A /* TenseEndingTable.swift in Sources */,
 				0A89DCC82BE11605CB866DEF /* TenseInfo.swift in Sources */,
 				46943ACFABF329DE1CBFC471 /* TensePill.swift in Sources */,
 				D7456B289D135CEB3A15122B /* TestResult.swift in Sources */,
+				81E4DB9F64F3FF3AB8BCB03A /* TextbookChapterListView.swift in Sources */,
+				B10A324C06F0957DDE2233F8 /* TextbookChapterView.swift in Sources */,
+				354631F309E625046A3A436B /* TextbookExerciseView.swift in Sources */,
 				27BA7FA9356467846A07697D /* TypingView.swift in Sources */,
 				943A94A8C71919F3EFC0E8FA /* UserProgress.swift in Sources */,
 				50E0095A23E119D1AB561232 /* VerbDetailView.swift in Sources */,
+				BA3DE2DA319AA3B572C19E11 /* VerbExampleCache.swift in Sources */,
+				4C577CF6B137D0A32759A169 /* VerbExampleGenerator.swift in Sources */,
 				81FA7EBCF18F0AAE0BF385C3 /* VerbListView.swift in Sources */,
 				4005E258FDF03C8B3A0D53BD /* VocabFlashcardView.swift in Sources */,
+				78FE99C5D511737B6877EDD5 /* VocabReviewView.swift in Sources */,
 				6BB4B0A655E6CB6F82D81B5A /* WeekTestView.swift in Sources */,
-				968D626462B0ADEC8D7D56AA /* CheckpointExamView.swift in Sources */,
 				E99473B7DF9BCAE150E9D1E1 /* WidgetDataService.swift in Sources */,
-							DDF58F3899FC4B92BF6587D2 /* StudyTimerService.swift in Sources */,
+			);
-				8C1E4E7F36D64EFF8D092AC8 /* StoryGenerator.swift in Sources */,
-				4C2649215B81470195F38ED0 /* StoryLibraryView.swift in Sources */,
-				8E3D8E8254CF4213B9D9FAD3 /* StoryReaderView.swift in Sources */,
-				12D2C9311D5C4764B48B1754 /* StoryQuizView.swift in Sources */,
-				8D7CA0F4496B44C28CD5EBD5 /* DictionaryService.swift in Sources */,
-				3EC2A2F4B9C24B029DA49C40 /* VocabReviewView.swift in Sources */,
-				53908E41767B438C8BD229CD /* ClozeView.swift in Sources */,
-				65ABC39F35804C619DAB3200 /* GrammarExercise.swift in Sources */,
-				B73F6EED00304B718C6FEFFA /* GrammarExerciseView.swift in Sources */,
-				EA07DB964C8940F69C14DE2C /* PronunciationService.swift in Sources */,
-				4DCC5CC233DE4701A12FD7EB /* ListeningView.swift in Sources */,
-				35D6404C60C249D5995AD895 /* ConversationService.swift in Sources */,
-				C8AF0931F7FD458C80B6EC0D /* ChatLibraryView.swift in Sources */,
-				6CCC8D51F5524688A4BC5AF8 /* ChatView.swift in Sources */,
-				8510085D78E248D885181E80 /* FeatureReferenceView.swift in Sources */,
-);
 			runOnlyForDeploymentPostprocessing = 0;
 		};
 		217A29BCEDD9D44B6DD85AF6 /* Sources */ = {

Conjuga/Conjuga/ConjugaApp.swift

@@ -40,6 +40,8 @@ struct ConjugaApp: App {
     @State private var syncMonitor = SyncStatusMonitor()
     @State private var studyTimer = StudyTimerService()
     @State private var dictionary = DictionaryService()
+    @State private var verbExampleCache = VerbExampleCache()
+    @State private var reflexiveStore = ReflexiveVerbStore()
 
     let localContainer: ModelContainer
     let cloudContainer: ModelContainer
@@ -69,12 +71,14 @@ struct ConjugaApp: App {
                 schema: Schema([
                     ReviewCard.self, CourseReviewCard.self, UserProgress.self,
                     TestResult.self, DailyLog.self, SavedSong.self, Story.self, Conversation.self,
+                    TextbookExerciseAttempt.self,
                 ]),
                 cloudKitDatabase: .private("iCloud.com.conjuga.app")
             )
             cloudContainer = try ModelContainer(
                 for: ReviewCard.self, CourseReviewCard.self, UserProgress.self,
                      TestResult.self, DailyLog.self, SavedSong.self, Story.self, Conversation.self,
+                     TextbookExerciseAttempt.self,
                 configurations: cloudConfig
             )
 
@@ -111,6 +115,8 @@ struct ConjugaApp: App {
             .environment(\.cloudModelContextProvider, { cloudContainer.mainContext })
             .environment(studyTimer)
             .environment(dictionary)
+            .environment(verbExampleCache)
+            .environment(reflexiveStore)
             .task {
                 let needsSeed = await DataLoader.needsSeeding(container: localContainer)
                 if needsSeed {
@@ -209,6 +215,7 @@ struct ConjugaApp: App {
             schema: Schema([
                 Verb.self, VerbForm.self, IrregularSpan.self,
                 TenseGuide.self, CourseDeck.self, VocabCard.self,
+                TextbookChapter.self,
             ]),
             url: url,
             cloudKitDatabase: .none
@@ -216,6 +223,7 @@ struct ConjugaApp: App {
         return try ModelContainer(
             for: Verb.self, VerbForm.self, IrregularSpan.self,
                  TenseGuide.self, CourseDeck.self, VocabCard.self,
+                 TextbookChapter.self,
             configurations: localConfig
         )
     }

Conjuga/Conjuga/Info.plist

@@ -2,6 +2,10 @@
 <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
 <plist version="1.0">
 <dict>
+	<key>BGTaskSchedulerPermittedIdentifiers</key>
+	<array>
+		<string>com.conjuga.app.refresh</string>
+	</array>
 	<key>CFBundleDevelopmentRegion</key>
 	<string>$(DEVELOPMENT_LANGUAGE)</string>
 	<key>CFBundleDisplayName</key>
@@ -22,21 +26,13 @@
 	<string>1</string>
 	<key>LSApplicationCategoryType</key>
 	<string>public.app-category.education</string>
-	<key>UILaunchScreen</key>
-	<dict/>
-	<key>NSSpeechRecognitionUsageDescription</key>
-	<string>Conjuga uses speech recognition to check your Spanish pronunciation and transcribe what you say during listening exercises.</string>
-	<key>NSMicrophoneUsageDescription</key>
-	<string>Conjuga needs microphone access to record your voice for pronunciation practice.</string>
 	<key>UIBackgroundModes</key>
 	<array>
 		<string>fetch</string>
 		<string>remote-notification</string>
 	</array>
-	<key>BGTaskSchedulerPermittedIdentifiers</key>
+	<key>UILaunchScreen</key>
-	<array>
+	<dict/>
-		<string>com.conjuga.app.refresh</string>
-	</array>
 	<key>UISupportedInterfaceOrientations</key>
 	<array>
 		<string>UIInterfaceOrientationPortrait</string>

Conjuga/Conjuga/Models/UserProgress.swift

@@ -14,6 +14,7 @@ final class UserProgress {
     var selectedLevel: String = "basic"
     var showVosotros: Bool = true
     var autoFillStem: Bool = false
+    var showReflexiveVerbsOnly: Bool = false
 
     // Legacy CloudKit array-backed fields retained for migration compatibility.
     var enabledTenses: [String] = []
@@ -21,6 +22,10 @@ final class UserProgress {
     var enabledTensesBlob: String = ""
     var unlockedBadgesBlob: String = ""
 
+    // Multi-select level + irregularity filters (Issue #26).
+    var selectedLevelsBlob: String = ""
+    var enabledIrregularCategoriesBlob: String = ""
+
     init() {}
 
     var selectedVerbLevel: VerbLevel {
@@ -44,6 +49,44 @@ final class UserProgress {
         }
     }
 
+    /// Levels currently enabled for practice. Multi-select per Issue #26.
+    /// Setting this also syncs `selectedLevel` to the highest-ranked selection so
+    /// legacy single-level consumers (widget, AI scenarios, word-of-day) stay consistent.
+    var selectedVerbLevels: Set<VerbLevel> {
+        get {
+            let raw = decodeStringArray(from: selectedLevelsBlob, fallback: [])
+            let decoded = Set(raw.compactMap(VerbLevel.init(rawValue:)))
+            if !decoded.isEmpty { return decoded }
+            // Pre-migration users: treat the single selectedLevel as the set.
+            if let legacy = VerbLevel(rawValue: selectedLevel) {
+                return [legacy]
+            }
+            return []
+        }
+        set {
+            let sorted = newValue.map(\.rawValue)
+            selectedLevelsBlob = Self.encodeStringArray(sorted)
+            selectedLevel = VerbLevel.highest(in: newValue)?.rawValue ?? VerbLevel.basic.rawValue
+        }
+    }
+
+    /// The single representative level for callers that need one value
+    /// (word-of-day widget, AI chat/story scenarios). Highest selected level.
+    var primaryLevel: VerbLevel {
+        VerbLevel.highest(in: selectedVerbLevels) ?? selectedVerbLevel
+    }
+
+    var enabledIrregularCategories: Set<IrregularSpan.SpanCategory> {
+        get {
+            let raw = decodeStringArray(from: enabledIrregularCategoriesBlob, fallback: [])
+            return Set(raw.compactMap(IrregularSpan.SpanCategory.init(rawValue:)))
+        }
+        set {
+            let sorted = newValue.map(\.rawValue)
+            enabledIrregularCategoriesBlob = Self.encodeStringArray(sorted)
+        }
+    }
+
     func setTenseEnabled(_ tenseId: String, enabled: Bool) {
         var values = Set(enabledTenseIDs)
         if enabled {
@@ -54,6 +97,26 @@ final class UserProgress {
         enabledTenseIDs = values.sorted()
     }
 
+    func setLevelEnabled(_ level: VerbLevel, enabled: Bool) {
+        var values = selectedVerbLevels
+        if enabled {
+            values.insert(level)
+        } else {
+            values.remove(level)
+        }
+        selectedVerbLevels = values
+    }
+
+    func setIrregularCategoryEnabled(_ category: IrregularSpan.SpanCategory, enabled: Bool) {
+        var values = enabledIrregularCategories
+        if enabled {
+            values.insert(category)
+        } else {
+            values.remove(category)
+        }
+        enabledIrregularCategories = values
+    }
+
     func unlockBadge(_ badgeId: String) {
         var values = Set(unlockedBadgeIDs)
         values.insert(badgeId)
@@ -67,6 +130,9 @@ final class UserProgress {
         if unlockedBadgesBlob.isEmpty && !unlockedBadges.isEmpty {
             unlockedBadgeIDs = unlockedBadges
         }
+        if selectedLevelsBlob.isEmpty, let legacy = VerbLevel(rawValue: selectedLevel) {
+            selectedVerbLevels = [legacy]
+        }
     }
 
     private func decodeStringArray(from blob: String, fallback: [String]) -> [String] {
@@ -86,4 +152,5 @@ final class UserProgress {
         }
         return string
     }
+
 }

Conjuga/Conjuga/Services/AnswerChecker.swift

@@ -0,0 +1,10 @@
+import Foundation
+import SharedModels
+
+/// Thin app-side wrapper around the SharedModels `AnswerGrader`. All logic
+/// lives in SharedModels so it can be unit tested.
+enum AnswerChecker {
+    static func grade(userText: String, canonical: String, alternates: [String] = []) -> TextbookGrade {
+        AnswerGrader.grade(userText: userText, canonical: canonical, alternates: alternates)
+    }
+}

Conjuga/Conjuga/Services/DataLoader.swift

@@ -6,6 +6,9 @@ actor DataLoader {
     static let courseDataVersion = 7
     static let courseDataKey = "courseDataVersion"
 
+    static let textbookDataVersion = 12
+    static let textbookDataKey = "textbookDataVersion"
+
     /// Quick check: does the DB need seeding or course data refresh?
     static func needsSeeding(container: ModelContainer) async -> Bool {
         let context = ModelContext(container)
@@ -15,6 +18,9 @@ actor DataLoader {
         let storedVersion = UserDefaults.standard.integer(forKey: courseDataKey)
         if storedVersion < courseDataVersion { return true }
 
+        let textbookVersion = UserDefaults.standard.integer(forKey: textbookDataKey)
+        if textbookVersion < textbookDataVersion { return true }
+
         return false
     }
 
@@ -133,6 +139,61 @@ actor DataLoader {
 
         // Seed course data (uses the same mainContext so @Query sees it)
         seedCourseData(context: context)
+
+        // Seed textbook data — only bump the version key if the seed
+        // actually inserted rows, so a missing/unparseable bundle doesn't
+        // permanently lock us out of future re-seeds.
+        if seedTextbookData(context: context) {
+            UserDefaults.standard.set(textbookDataVersion, forKey: textbookDataKey)
+        }
+    }
+
+    /// Re-seed textbook data if the version has changed OR if the rows are
+    /// missing on disk. The row-count check exists because anything opening
+    /// this store with a subset schema (e.g. an out-of-date widget extension)
+    /// can destructively drop the rows without touching UserDefaults — so a
+    /// pure version-flag trigger would leave us permanently empty.
+    static func refreshTextbookDataIfNeeded(container: ModelContainer) async {
+        let shared = UserDefaults.standard
+        let context = ModelContext(container)
+        let existingCount = (try? context.fetchCount(FetchDescriptor<TextbookChapter>())) ?? 0
+        let versionCurrent = shared.integer(forKey: textbookDataKey) >= textbookDataVersion
+
+        if versionCurrent && existingCount > 0 { return }
+
+        if versionCurrent {
+            print("Textbook data version current but store has \(existingCount) chapters — re-seeding...")
+        } else {
+            print("Textbook data version outdated — re-seeding...")
+        }
+
+        // Fetch + delete individually instead of batch delete. SwiftData's
+        // context.delete(model:) hits the store directly and doesn't always
+        // clear the unique-constraint index before the reseed's save runs,
+        // so re-inserting rows with the same .unique id can throw.
+        let textbookCourseName = "Complete Spanish Step-by-Step"
+        if let existing = try? context.fetch(FetchDescriptor<TextbookChapter>()) {
+            for chapter in existing { context.delete(chapter) }
+        }
+        let deckDescriptor = FetchDescriptor<CourseDeck>(
+            predicate: #Predicate<CourseDeck> { $0.courseName == textbookCourseName }
+        )
+        if let decks = try? context.fetch(deckDescriptor) {
+            for deck in decks { context.delete(deck) }
+        }
+        do {
+            try context.save()
+        } catch {
+            print("[DataLoader] ERROR: textbook wipe save failed: \(error)")
+            return
+        }
+
+        if seedTextbookData(context: context) {
+            shared.set(textbookDataVersion, forKey: textbookDataKey)
+            print("Textbook data re-seeded to version \(textbookDataVersion)")
+        } else {
+            print("Textbook re-seed failed — leaving version key untouched so next launch retries")
+        }
     }
 
     /// Re-seed course data if the version has changed (e.g. examples were added).
@@ -170,6 +231,10 @@ actor DataLoader {
         // Re-seed course data
         seedCourseData(context: context)
 
+        // Textbook's vocab decks/cards share the same CourseDeck/VocabCard
+        // entities, so they were just wiped above. Reseed them.
+        seedTextbookVocabDecks(context: context, courseName: "Complete Spanish Step-by-Step")
+
         shared.set(courseDataVersion, forKey: courseDataKey)
         print("Course data re-seeded to version \(courseDataVersion)")
     }
@@ -336,4 +401,169 @@ actor DataLoader {
         context.insert(reviewCard)
         return reviewCard
     }
+
+    // MARK: - Textbook seeding
+
+    @discardableResult
+    private static func seedTextbookData(context: ModelContext) -> Bool {
+        let url = Bundle.main.url(forResource: "textbook_data", withExtension: "json")
+            ?? Bundle.main.bundleURL.appendingPathComponent("textbook_data.json")
+        guard let data = try? Data(contentsOf: url) else {
+            print("[DataLoader] textbook_data.json not bundled — skipping textbook seed")
+            return false
+        }
+        guard let json = try? JSONSerialization.jsonObject(with: data) as? [String: Any] else {
+            print("[DataLoader] ERROR: Could not parse textbook_data.json")
+            return false
+        }
+        let courseName = (json["courseName"] as? String) ?? "Textbook"
+        guard let chapters = json["chapters"] as? [[String: Any]] else {
+            print("[DataLoader] ERROR: textbook_data.json missing chapters")
+            return false
+        }
+
+        var inserted = 0
+        for ch in chapters {
+            guard let id = ch["id"] as? String,
+                  let number = ch["number"] as? Int,
+                  let title = ch["title"] as? String,
+                  let blocksRaw = ch["blocks"] as? [[String: Any]] else { continue }
+
+            let part = (ch["part"] as? Int) ?? 0
+
+            // Normalize each block to canonical keys expected by TextbookBlock decoder.
+            var normalized: [[String: Any]] = []
+            var exerciseCount = 0
+            var vocabTableCount = 0
+            for (i, b) in blocksRaw.enumerated() {
+                var out: [String: Any] = [:]
+                out["index"] = i
+                let kind = (b["kind"] as? String) ?? ""
+                out["kind"] = kind
+                switch kind {
+                case "heading":
+                    if let level = b["level"] { out["level"] = level }
+                    if let text = b["text"] { out["text"] = text }
+                case "paragraph":
+                    if let text = b["text"] { out["text"] = text }
+                case "key_vocab_header":
+                    break
+                case "vocab_table":
+                    vocabTableCount += 1
+                    if let src = b["sourceImage"] { out["sourceImage"] = src }
+                    if let lines = b["ocrLines"] { out["ocrLines"] = lines }
+                    if let conf = b["ocrConfidence"] { out["ocrConfidence"] = conf }
+                    // Paired Spanish→English cards from the bounding-box extractor.
+                    if let cards = b["cards"] as? [[String: Any]], !cards.isEmpty {
+                        let normalized: [[String: Any]] = cards.compactMap { c in
+                            guard let front = c["front"] as? String,
+                                  let back = c["back"] as? String else { return nil }
+                            return ["front": front, "back": back]
+                        }
+                        if !normalized.isEmpty {
+                            out["cards"] = normalized
+                        }
+                    }
+                case "exercise":
+                    exerciseCount += 1
+                    if let exId = b["id"] { out["exerciseId"] = exId }
+                    if let inst = b["instruction"] { out["instruction"] = inst }
+                    if let extra = b["extra"] { out["extra"] = extra }
+                    if let prompts = b["prompts"] { out["prompts"] = prompts }
+                    if let items = b["answerItems"] { out["answerItems"] = items }
+                    if let freeform = b["freeform"] { out["freeform"] = freeform }
+                default:
+                    break
+                }
+                normalized.append(out)
+            }
+
+            let bodyJSON: Data
+            do {
+                bodyJSON = try JSONSerialization.data(withJSONObject: normalized, options: [])
+            } catch {
+                print("[DataLoader] failed to encode chapter \(number) blocks: \(error)")
+                continue
+            }
+
+            let chapter = TextbookChapter(
+                id: id,
+                number: number,
+                title: title,
+                part: part,
+                courseName: courseName,
+                bodyJSON: bodyJSON,
+                exerciseCount: exerciseCount,
+                vocabTableCount: vocabTableCount
+            )
+            context.insert(chapter)
+            inserted += 1
+        }
+
+        do {
+            try context.save()
+        } catch {
+            print("[DataLoader] ERROR: textbook chapter save failed: \(error)")
+            return false
+        }
+
+        // Verify rows actually hit the store — guards against the case where
+        // save returned cleanly but no rows were persisted.
+        let persisted = (try? context.fetchCount(FetchDescriptor<TextbookChapter>())) ?? 0
+        guard persisted > 0 else {
+            print("[DataLoader] ERROR: textbook seeded \(inserted) chapters but persisted count is 0")
+            return false
+        }
+
+        // Seed textbook-derived vocabulary flashcards as CourseDecks so the
+        // existing Course UI can surface them alongside LanGo decks.
+        seedTextbookVocabDecks(context: context, courseName: courseName)
+
+        print("Textbook seeding complete: \(inserted) chapters inserted, \(persisted) persisted")
+        return true
+    }
+
+    private static func seedTextbookVocabDecks(context: ModelContext, courseName: String) {
+        let url = Bundle.main.url(forResource: "textbook_vocab", withExtension: "json")
+            ?? Bundle.main.bundleURL.appendingPathComponent("textbook_vocab.json")
+        guard let data = try? Data(contentsOf: url),
+              let json = try? JSONSerialization.jsonObject(with: data) as? [String: Any],
+              let chaptersArr = json["chapters"] as? [[String: Any]]
+        else { return }
+
+        let courseSlug = courseName.lowercased()
+            .replacingOccurrences(of: " ", with: "-")
+
+        var deckCount = 0
+        var cardCount = 0
+        for chData in chaptersArr {
+            guard let chNum = chData["chapter"] as? Int,
+                  let cards = chData["cards"] as? [[String: Any]],
+                  !cards.isEmpty else { continue }
+
+            let deckId = "textbook_\(courseSlug)_ch\(chNum)"
+            let title = "Chapter \(chNum) vocabulary"
+            let deck = CourseDeck(
+                id: deckId,
+                weekNumber: chNum,
+                title: title,
+                cardCount: cards.count,
+                courseName: courseName,
+                isReversed: false
+            )
+            context.insert(deck)
+            deckCount += 1
+
+            for c in cards {
+                guard let front = c["front"] as? String,
+                      let back = c["back"] as? String else { continue }
+                let card = VocabCard(front: front, back: back, deckId: deckId)
+                card.deck = deck
+                context.insert(card)
+                cardCount += 1
+            }
+        }
+        try? context.save()
+        print("Textbook vocab seeding complete: \(deckCount) decks, \(cardCount) cards")
+    }
 }

Conjuga/Conjuga/Services/PracticeSessionService.swift

@@ -4,14 +4,23 @@ import SwiftData
 
 struct PracticeSettings: Sendable {
     let selectedLevel: String
+    let selectedLevels: Set<String>
     let enabledTenses: Set<String>
+    let enabledIrregularCategories: Set<IrregularSpan.SpanCategory>
     let showVosotros: Bool
+    let showReflexiveVerbsOnly: Bool
+    let reflexiveBaseInfinitives: Set<String>
 
-    init(progress: UserProgress?) {
+    init(progress: UserProgress?, reflexiveBaseInfinitives: Set<String> = []) {
-        let resolved = progress?.enabledTenseIDs ?? []
+        let resolvedTenses = progress?.enabledTenseIDs ?? []
+        let resolvedLevels = progress?.selectedVerbLevels ?? []
         self.selectedLevel = progress?.selectedLevel ?? VerbLevel.basic.rawValue
-        self.enabledTenses = Set(resolved)
+        self.selectedLevels = Set(resolvedLevels.map(\.rawValue))
+        self.enabledTenses = Set(resolvedTenses)
+        self.enabledIrregularCategories = progress?.enabledIrregularCategories ?? []
         self.showVosotros = progress?.showVosotros ?? true
+        self.showReflexiveVerbsOnly = progress?.showReflexiveVerbsOnly ?? false
+        self.reflexiveBaseInfinitives = reflexiveBaseInfinitives
     }
 
     var selectionTenseIDs: [String] {
@@ -36,16 +45,25 @@ struct FullTablePrompt {
 struct PracticeSessionService {
     let localContext: ModelContext
     let cloudContext: ModelContext
+    let reflexiveBaseInfinitives: Set<String>
     private let referenceStore: ReferenceStore
 
-    init(localContext: ModelContext, cloudContext: ModelContext) {
+    init(
+        localContext: ModelContext,
+        cloudContext: ModelContext,
+        reflexiveBaseInfinitives: Set<String> = []
+    ) {
         self.localContext = localContext
         self.cloudContext = cloudContext
+        self.reflexiveBaseInfinitives = reflexiveBaseInfinitives
         self.referenceStore = ReferenceStore(context: localContext)
     }
 
     func settings() -> PracticeSettings {
-        PracticeSettings(progress: ReviewStore.fetchOrCreateUserProgress(context: cloudContext))
+        PracticeSettings(
+            progress: ReviewStore.fetchOrCreateUserProgress(context: cloudContext),
+            reflexiveBaseInfinitives: reflexiveBaseInfinitives
+        )
     }
 
     func nextCard(for focusMode: FocusMode, excludingVerbId lastVerbId: Int? = nil) -> PracticeCardLoad? {
@@ -79,7 +97,12 @@ struct PracticeSessionService {
 
     func randomFullTablePrompt() -> FullTablePrompt? {
         let settings = settings()
-        let verbs = referenceStore.fetchVerbs(selectedLevel: settings.selectedLevel)
+        // Full Table practice is regular-only, so the irregular-category setting is
+        // deliberately ignored here (applying it would empty the pool).
+        let verbs = applyReflexiveFilter(
+            to: referenceStore.fetchVerbs(selectedLevels: settings.selectedLevels),
+            settings: settings
+        )
         guard !verbs.isEmpty else { return nil }
 
         for _ in 0..<40 {
@@ -89,6 +112,11 @@ struct PracticeSessionService {
 
             let forms = referenceStore.fetchForms(verbId: verb.id, tenseId: tenseId)
             if forms.isEmpty { continue }
+
+            // Full Table practice is for regular patterns only — skip combos
+            // where any form in this (verb, tense) is irregular.
+            if forms.contains(where: { $0.regularity != "regular" }) { continue }
+
             return FullTablePrompt(verb: verb, tenseInfo: tenseInfo, forms: forms)
         }
 
@@ -131,6 +159,27 @@ struct PracticeSessionService {
         return buildCardLoad(verb: verb, form: form)
     }
 
+    /// When the user has "Reflexive verbs only" enabled, restrict the allowed
+    /// verb-id set to IDs whose infinitive is in the curated list.
+    /// No-op otherwise.
+    private func applyReflexiveFilter(to ids: Set<Int>, settings: PracticeSettings) -> Set<Int> {
+        guard settings.showReflexiveVerbsOnly, !settings.reflexiveBaseInfinitives.isEmpty else {
+            return ids
+        }
+        let matching = ids.filter { id in
+            guard let verb = referenceStore.fetchVerb(id: id) else { return false }
+            return settings.reflexiveBaseInfinitives.contains(verb.infinitive.lowercased())
+        }
+        return matching
+    }
+
+    private func applyReflexiveFilter(to verbs: [Verb], settings: PracticeSettings) -> [Verb] {
+        guard settings.showReflexiveVerbsOnly, !settings.reflexiveBaseInfinitives.isEmpty else {
+            return verbs
+        }
+        return verbs.filter { settings.reflexiveBaseInfinitives.contains($0.infinitive.lowercased()) }
+    }
+
     private func buildCardLoad(verb: Verb, form: VerbForm) -> PracticeCardLoad {
         let spans = referenceStore.fetchSpans(
             verbId: form.verbId,
@@ -152,7 +201,13 @@ struct PracticeSessionService {
 
     private func fetchDueCard(excluding lastVerbId: Int?) -> ReviewCard? {
         let settings = settings()
-        let allowedVerbIds = referenceStore.allowedVerbIDs(selectedLevel: settings.selectedLevel)
+        let allowedVerbIds = applyReflexiveFilter(
+            to: referenceStore.allowedVerbIDs(
+                selectedLevels: settings.selectedLevels,
+                irregularCategories: settings.enabledIrregularCategories
+            ),
+            settings: settings
+        )
         let now = Date()
         var descriptor = FetchDescriptor<ReviewCard>(
             predicate: #Predicate<ReviewCard> { $0.dueDate <= now },
@@ -179,7 +234,13 @@ struct PracticeSessionService {
 
     private func pickWeakForm() -> VerbForm? {
         let settings = settings()
-        let allowedVerbIds = referenceStore.allowedVerbIDs(selectedLevel: settings.selectedLevel)
+        let allowedVerbIds = applyReflexiveFilter(
+            to: referenceStore.allowedVerbIDs(
+                selectedLevels: settings.selectedLevels,
+                irregularCategories: settings.enabledIrregularCategories
+            ),
+            settings: settings
+        )
 
         let descriptor = FetchDescriptor<ReviewCard>(
             predicate: #Predicate<ReviewCard> { $0.easeFactor < 2.0 && $0.repetitions > 0 },
@@ -201,7 +262,15 @@ struct PracticeSessionService {
 
     private func pickIrregularForm(filter: IrregularityFilter) -> VerbForm? {
         let settings = settings()
-        let allowedVerbIds = referenceStore.allowedVerbIDs(selectedLevel: settings.selectedLevel)
+        // Focus mode explicitly selects one irregular category, so the user's
+        // settings-level irregular filter is deliberately skipped here.
+        let allowedVerbIds = applyReflexiveFilter(
+            to: referenceStore.allowedVerbIDs(
+                selectedLevels: settings.selectedLevels,
+                irregularCategories: []
+            ),
+            settings: settings
+        )
         let typeRange: ClosedRange<Int>
 
         switch filter {
@@ -238,7 +307,13 @@ struct PracticeSessionService {
     private func pickCommonTenseForm() -> VerbForm? {
         let settings = settings()
         let coreTenseIDs = TenseID.coreTenseIDs
-        let verbs = referenceStore.fetchVerbs(selectedLevel: settings.selectedLevel)
+        let verbs = applyReflexiveFilter(
+            to: referenceStore.fetchVerbs(
+                selectedLevels: settings.selectedLevels,
+                irregularCategories: settings.enabledIrregularCategories
+            ),
+            settings: settings
+        )
         guard let verb = verbs.randomElement() else { return nil }
 
         let forms = referenceStore.fetchVerbForms(verbId: verb.id).filter { form in
@@ -251,7 +326,13 @@ struct PracticeSessionService {
 
     private func pickRandomForm() -> VerbForm? {
         let settings = settings()
-        let verbs = referenceStore.fetchVerbs(selectedLevel: settings.selectedLevel)
+        let verbs = applyReflexiveFilter(
+            to: referenceStore.fetchVerbs(
+                selectedLevels: settings.selectedLevels,
+                irregularCategories: settings.enabledIrregularCategories
+            ),
+            settings: settings
+        )
         guard let verb = verbs.randomElement() else { return nil }
 
         let forms = referenceStore.fetchVerbForms(verbId: verb.id).filter { form in

Conjuga/Conjuga/Services/PronunciationService.swift

@@ -65,28 +65,26 @@ final class PronunciationService {
 
         do {
             let audioSession = AVAudioSession.sharedInstance()
-            try audioSession.setCategory(.playAndRecord, mode: .measurement, options: [.duckOthers, .defaultToSpeaker])
+            try audioSession.setCategory(.record, mode: .measurement, options: [.duckOthers])
             try audioSession.setActive(true, options: .notifyOthersOnDeactivation)
 
-            audioEngine = AVAudioEngine()
+            // Use SFSpeechAudioBufferRecognitionRequest with the recognizer
+            // directly — avoid AVAudioEngine entirely since it produces
+            // zero-length buffers on some devices causing assertion crashes.
             request = SFSpeechAudioBufferRecognitionRequest()
-
+            guard let request else { return }
-            guard let audioEngine, let request else { return }
             request.shouldReportPartialResults = true
+            request.requiresOnDeviceRecognition = recognizer.supportsOnDeviceRecognition
+
+            // Use AVAudioEngine with the native input format
+            audioEngine = AVAudioEngine()
+            guard let audioEngine else { return }
 
             let inputNode = audioEngine.inputNode
-            let recordingFormat = inputNode.outputFormat(forBus: 0)
 
-            // Validate format — 0 channels crashes installTap
+            // Use nil format — lets the system pick a compatible format
-            guard recordingFormat.channelCount > 0 else {
+            // and avoids the mDataByteSize(0) assertion from format mismatches
-                print("[PronunciationService] invalid recording format (0 channels)")
+            inputNode.installTap(onBus: 0, bufferSize: 4096, format: nil) { buffer, _ in
-                self.audioEngine = nil
-                self.request = nil
-                return
-            }
-
-            inputNode.installTap(onBus: 0, bufferSize: 1024, format: recordingFormat) { buffer, _ in
-                guard buffer.frameLength > 0 else { return }
                 request.append(buffer)
             }
 

Conjuga/Conjuga/Services/ReferenceStore.swift

@@ -27,6 +27,50 @@ struct ReferenceStore {
         Set(fetchVerbs(selectedLevel: selectedLevel).map(\.id))
     }
 
+    /// Union of data-levels for all selected user-facing levels.
+    /// Empty input produces an empty result — callers decide how to handle that.
+    func fetchVerbs(selectedLevels: Set<String>) -> [Verb] {
+        guard !selectedLevels.isEmpty else { return [] }
+        let ids = PracticeFilter.verbIDs(
+            matchingLevels: selectedLevels,
+            in: fetchVerbs().map { .init(id: $0.id, level: $0.level) }
+        )
+        return fetchVerbs().filter { ids.contains($0.id) }
+    }
+
+    /// Practice verb pool intersecting selected levels with selected irregular-span categories.
+    /// Delegates to `PracticeFilter` so the intersection logic is unit-tested
+    /// in SharedModels without a ModelContainer (Issue #26).
+    func allowedVerbIDs(
+        selectedLevels: Set<String>,
+        irregularCategories: Set<IrregularSpan.SpanCategory>
+    ) -> Set<Int> {
+        PracticeFilter.allowedVerbIDs(
+            verbs: fetchVerbs().map { .init(id: $0.id, level: $0.level) },
+            spans: allIrregularSlots(),
+            selectedLevels: selectedLevels,
+            irregularCategories: irregularCategories
+        )
+    }
+
+    /// Convenience: full Verb objects passing both filters.
+    func fetchVerbs(
+        selectedLevels: Set<String>,
+        irregularCategories: Set<IrregularSpan.SpanCategory>
+    ) -> [Verb] {
+        let ids = allowedVerbIDs(
+            selectedLevels: selectedLevels,
+            irregularCategories: irregularCategories
+        )
+        return fetchVerbs().filter { ids.contains($0.id) }
+    }
+
+    private func allIrregularSlots() -> [PracticeFilter.IrregularSlot] {
+        let descriptor = FetchDescriptor<IrregularSpan>()
+        let spans = (try? context.fetch(descriptor)) ?? []
+        return spans.map { .init(verbId: $0.verbId, category: $0.category) }
+    }
+
     func fetchVerb(id: Int) -> Verb? {
         let descriptor = FetchDescriptor<Verb>(predicate: #Predicate<Verb> { $0.id == id })
         return (try? context.fetch(descriptor))?.first

Conjuga/Conjuga/Services/ReflexiveVerbStore.swift

@@ -0,0 +1,59 @@
+import Foundation
+import SharedModels
+
+/// Loads and queries the curated reflexive-verb list bundled with the app
+/// (Gitea issue #28). One JSON load at init; in-memory lookup thereafter.
+///
+/// `entries(for:)` returns a list because a single base infinitive may map to
+/// multiple reflexive entries — e.g., `ponerse` covers both "to put on
+/// (clothing) / to become" and "to come to an agreement (with)".
+@MainActor
+@Observable
+final class ReflexiveVerbStore {
+
+    /// Process-wide accessor for services that can't use @Environment injection
+    /// (e.g. PracticeSessionService called from ViewModels). Views should still
+    /// prefer @Environment(ReflexiveVerbStore.self) for consistency.
+    static let shared = ReflexiveVerbStore()
+
+    private(set) var entries: [ReflexiveVerb] = []
+    private var indexByBase: [String: [ReflexiveVerb]] = [:]
+
+    /// Set of base infinitives present in the list. Cheap lookup for filters.
+    private(set) var baseInfinitives: Set<String> = []
+
+    init(bundle: Bundle = .main) {
+        load(from: bundle)
+    }
+
+    /// All reflexive entries whose base infinitive matches (case-insensitive).
+    func entries(for baseInfinitive: String) -> [ReflexiveVerb] {
+        indexByBase[baseInfinitive.lowercased()] ?? []
+    }
+
+    /// Convenience — true when the verb's bare infinitive appears in the list.
+    func isReflexive(baseInfinitive: String) -> Bool {
+        baseInfinitives.contains(baseInfinitive.lowercased())
+    }
+
+    private func load(from bundle: Bundle) {
+        guard let url = bundle.url(forResource: "reflexive_verbs", withExtension: "json"),
+              let data = try? Data(contentsOf: url) else {
+            print("[ReflexiveVerbStore] bundled reflexive_verbs.json not found")
+            return
+        }
+        do {
+            let decoded = try JSONDecoder().decode([ReflexiveVerb].self, from: data)
+            entries = decoded
+            var index: [String: [ReflexiveVerb]] = [:]
+            for entry in decoded {
+                index[entry.baseInfinitive.lowercased(), default: []].append(entry)
+            }
+            indexByBase = index
+            baseInfinitives = Set(index.keys)
+            print("[ReflexiveVerbStore] loaded \(decoded.count) entries (\(baseInfinitives.count) distinct base infinitives)")
+        } catch {
+            print("[ReflexiveVerbStore] decode failed: \(error)")
+        }
+    }
+}

Conjuga/Conjuga/Services/StartupCoordinator.swift

@@ -9,6 +9,7 @@ enum StartupCoordinator {
     static func bootstrap(localContainer: ModelContainer) async {
         await DataLoader.seedIfNeeded(container: localContainer)
         await DataLoader.refreshCourseDataIfNeeded(container: localContainer)
+        await DataLoader.refreshTextbookDataIfNeeded(container: localContainer)
     }
 
     /// Recurring maintenance: legacy migrations, identity repair, cloud dedup.

Conjuga/Conjuga/Services/StoreInspector.swift

@@ -26,12 +26,14 @@ enum StoreInspector {
         let hasZVERBFORM = tables.contains("ZVERBFORM")
         let hasZTENSEGUIDE = tables.contains("ZTENSEGUIDE")
         let hasZVOCABCARD = tables.contains("ZVOCABCARD")
+        let hasZTEXTBOOKCHAPTER = tables.contains("ZTEXTBOOKCHAPTER")
 
         var summary = "[StoreInspector:\(label)] \(tables.count) tables"
         summary += " | ZVERB=\(hasZVERB ? queryInt(db: db, sql: "SELECT COUNT(*) FROM ZVERB") : -1)"
         summary += " ZVERBFORM=\(hasZVERBFORM ? queryInt(db: db, sql: "SELECT COUNT(*) FROM ZVERBFORM") : -1)"
         summary += " ZTENSEGUIDE=\(hasZTENSEGUIDE ? queryInt(db: db, sql: "SELECT COUNT(*) FROM ZTENSEGUIDE") : -1)"
         summary += " ZVOCABCARD=\(hasZVOCABCARD ? queryInt(db: db, sql: "SELECT COUNT(*) FROM ZVOCABCARD") : -1)"
+        summary += " ZTEXTBOOKCHAPTER=\(hasZTEXTBOOKCHAPTER ? queryInt(db: db, sql: "SELECT COUNT(*) FROM ZTEXTBOOKCHAPTER") : -1)"
         print(summary)
 
         // Log all Z-tables (SwiftData entity tables start with Z, minus Core Data system tables)

Conjuga/Conjuga/Services/VerbExampleCache.swift

@@ -0,0 +1,65 @@
+import Foundation
+import SharedModels
+
+/// Disk-backed cache for verb example sentences (Issue #27). One JSON file
+/// in the Caches directory keyed by verb id; lazy-loaded on first access and
+/// write-through on every generation. Matches DictionaryService's disk pattern.
+///
+/// Cache eviction by the OS is acceptable because contents are regenerable.
+@MainActor
+@Observable
+final class VerbExampleCache {
+
+    private var store: [Int: [VerbExample]] = [:]
+    private var isLoaded = false
+
+    init() {}
+
+    // MARK: - Public API
+
+    /// Look up cached examples for a verb; returns nil on miss.
+    /// Safe to call before `loadIfNeeded()`; it triggers the disk load itself.
+    func examples(for verbId: Int) -> [VerbExample]? {
+        loadIfNeeded()
+        return store[verbId]
+    }
+
+    /// Store newly generated examples and persist to disk.
+    func setExamples(_ examples: [VerbExample], for verbId: Int) {
+        loadIfNeeded()
+        store[verbId] = examples
+        save()
+    }
+
+    // MARK: - Disk I/O
+
+    private static var cacheURL: URL {
+        FileManager.default.urls(for: .cachesDirectory, in: .userDomainMask)[0]
+            .appendingPathComponent("verb_examples.json")
+    }
+
+    private func loadIfNeeded() {
+        guard !isLoaded else { return }
+        defer { isLoaded = true }
+
+        guard let data = try? Data(contentsOf: Self.cacheURL),
+              let decoded = try? JSONDecoder().decode([String: [VerbExample]].self, from: data)
+        else { return }
+
+        // Persisted with String keys because JSON object keys are strings;
+        // convert back to Int for in-memory lookup.
+        var rebuilt: [Int: [VerbExample]] = [:]
+        for (key, value) in decoded {
+            if let id = Int(key) {
+                rebuilt[id] = value
+            }
+        }
+        store = rebuilt
+    }
+
+    private func save() {
+        let serialized = Dictionary(uniqueKeysWithValues: store.map { (String($0.key), $0.value) })
+        guard let data = try? JSONEncoder().encode(serialized) else { return }
+        try? data.write(to: Self.cacheURL)
+    }
+}

Conjuga/Conjuga/Services/VerbExampleGenerator.swift

@@ -0,0 +1,78 @@
+import Foundation
+import FoundationModels
+import SharedModels
+
+/// Generates a set of example sentences for a single verb, one per core tense
+/// (Issue #27). Mirrors the StoryGenerator pattern: @Generable response types,
+/// a static availability flag, and a single generate(...) entry point.
+@MainActor
+struct VerbExampleGenerator {
+
+    // MARK: - Generable Types
+
+    @Generable
+    struct GeneratedExampleSet {
+        @Guide(
+            description: "Six example sentences, one per tense in the exact order requested. Each sentence must actually use the target verb conjugated in that tense.",
+            .count(6)
+        )
+        var examples: [GeneratedExample]
+    }
+
+    @Generable
+    struct GeneratedExample {
+        @Guide(description: "The tense id this sentence demonstrates. Must match one of the ids provided in the prompt exactly (e.g. ind_presente).")
+        var tenseId: String
+
+        @Guide(description: "A natural Spanish sentence, 6-14 words, that uses the target verb in the specified tense. For imperative tenses use tú or nosotros — never yo.")
+        var spanish: String
+
+        @Guide(description: "An accurate, idiomatic English translation of the Spanish sentence.")
+        var english: String
+    }
+
+    // MARK: - Generation
+
+    /// Generate one example per tense in `tenseIds`. Returns the examples in the
+    /// same order as `tenseIds`, filling in placeholders for any the model skipped.
+    static func generate(
+        verbInfinitive: String,
+        verbEnglish: String,
+        tenseIds: [String]
+    ) async throws -> [VerbExample] {
+        let tenseList = tenseIds
+            .compactMap { id in TenseInfo.find(id).map { "\(id) (\($0.english))" } }
+            .joined(separator: ", ")
+
+        let session = LanguageModelSession(instructions: """
+            You are a Spanish language teacher writing short example sentences for a learner.
+            The learner is studying the verb "\(verbInfinitive)" (to \(verbEnglish)).
+            Write one sentence per requested tense. Each sentence must:
+            - Actually conjugate "\(verbInfinitive)" in that tense (not just mention it).
+            - Be 6-14 words, natural and everyday.
+            - Use vocabulary appropriate for intermediate learners.
+            - Vary subjects and contexts across the set; do not reuse the same subject twice.
+            For imperative tenses, address "tú" or "nosotros" — never "yo".
+            """)
+
+        let prompt = """
+            Write example sentences for "\(verbInfinitive)" in these tenses, in this order:
+            \(tenseList)
+
+            Return one GeneratedExample per tense with the matching tenseId, spanish, and english.
+            """
+
+        let response = try await session.respond(to: prompt, generating: GeneratedExampleSet.self)
+
+        // Map by tenseId and return in the caller's requested order so the UI
+        // renders a predictable sequence even if the model shuffles its output.
+        let byTense = Dictionary(uniqueKeysWithValues: response.content.examples.map {
+            ($0.tenseId, VerbExample(tenseId: $0.tenseId, spanish: $0.spanish, english: $0.english))
+        })
+        return tenseIds.compactMap { byTense[$0] }
+    }
+
+    static var isAvailable: Bool {
+        SystemLanguageModel.default.availability == .available
+    }
+}

Conjuga/Conjuga/ViewModels/PracticeViewModel.swift

@@ -96,7 +96,11 @@ final class PracticeViewModel {
         currentSpans = []
         hasCards = true
         isLoading = true
-        let service = PracticeSessionService(localContext: localContext, cloudContext: cloudContext)
+        let service = PracticeSessionService(
+            localContext: localContext,
+            cloudContext: cloudContext,
+            reflexiveBaseInfinitives: ReflexiveVerbStore.shared.baseInfinitives
+        )
         guard let cardLoad = service.nextCard(for: focusMode, excludingVerbId: currentVerb?.id) else {
             clearCurrentCard()
             hasCards = false

Conjuga/Conjuga/Views/Course/CourseView.swift

@@ -5,9 +5,14 @@ import SwiftData
 struct CourseView: View {
     @Environment(\.cloudModelContextProvider) private var cloudModelContextProvider
     @Query(sort: \CourseDeck.weekNumber) private var decks: [CourseDeck]
+    @Query(sort: \TextbookChapter.number) private var textbookChapters: [TextbookChapter]
     @AppStorage("selectedCourse") private var selectedCourse: String?
     @State private var testResults: [TestResult] = []
 
+    private var textbookCourses: [String] {
+        Array(Set(textbookChapters.map(\.courseName))).sorted()
+    }
+
     private var cloudModelContext: ModelContext { cloudModelContextProvider() }
 
     private var courseNames: [String] {
@@ -62,6 +67,32 @@ struct CourseView: View {
                         description: Text("Course data is loading...")
                     )
                 } else {
+                    // Textbook entry (shown above course picker when available)
+                    if !textbookCourses.isEmpty {
+                        Section {
+                            ForEach(textbookCourses, id: \.self) { name in
+                                NavigationLink(value: TextbookDestination(courseName: name)) {
+                                    HStack(spacing: 12) {
+                                        Image(systemName: "book.fill")
+                                            .font(.title3)
+                                            .foregroundStyle(.indigo)
+                                            .frame(width: 32)
+                                        VStack(alignment: .leading, spacing: 2) {
+                                            Text(name)
+                                                .font(.subheadline.weight(.semibold))
+                                            Text("Read chapters, do exercises")
+                                                .font(.caption)
+                                                .foregroundStyle(.secondary)
+                                        }
+                                        Spacer()
+                                    }
+                                }
+                            }
+                        } header: {
+                            Text("Textbook")
+                        }
+                    }
+
                     // Course picker
                     if courseNames.count > 1 {
                         Section {
@@ -155,6 +186,24 @@ struct CourseView: View {
             .navigationDestination(for: CheckpointDestination.self) { dest in
                 CheckpointExamView(courseName: dest.courseName, throughWeek: dest.throughWeek)
             }
+            .navigationDestination(for: TextbookDestination.self) { dest in
+                TextbookChapterListView(courseName: dest.courseName)
+            }
+            .navigationDestination(for: TextbookChapter.self) { chapter in
+                TextbookChapterView(chapter: chapter)
+            }
+            .navigationDestination(for: TextbookExerciseDestination.self) { dest in
+                textbookExerciseView(for: dest)
+            }
+        }
+    }
+
+    @ViewBuilder
+    private func textbookExerciseView(for dest: TextbookExerciseDestination) -> some View {
+        if let chapter = textbookChapters.first(where: { $0.id == dest.chapterId }) {
+            TextbookExerciseView(chapter: chapter, blockIndex: dest.blockIndex)
+        } else {
+            ContentUnavailableView("Exercise unavailable", systemImage: "questionmark.circle")
         }
     }
 
@@ -175,6 +224,10 @@ struct CheckpointDestination: Hashable {
     let throughWeek: Int
 }
 
+struct TextbookDestination: Hashable {
+    let courseName: String
+}
+
 // MARK: - Deck Row
 
 private struct DeckRowView: View {

Conjuga/Conjuga/Views/Course/DeckStudyView.swift

@@ -8,6 +8,11 @@ struct DeckStudyView: View {
     @State private var isStudying = false
     @State private var speechService = SpeechService()
     @State private var deckCards: [VocabCard] = []
+    @State private var expandedConjugations: Set<String> = []
+
+    private var isStemChangingDeck: Bool {
+        deck.title.localizedCaseInsensitiveContains("stem changing")
+    }
 
     var body: some View {
         cardListView
@@ -19,7 +24,8 @@ struct DeckStudyView: View {
                     VocabFlashcardView(
                         cards: deckCards.shuffled(),
                         speechService: speechService,
-                        onDone: { isStudying = false }
+                        onDone: { isStudying = false },
+                        deckTitle: deck.title
                     )
                     .toolbar {
                         ToolbarItem(placement: .cancellationAction) {
@@ -30,6 +36,24 @@ struct DeckStudyView: View {
             }
     }
 
+    /// Reversed stem-change decks have `front` as English, so prefer the
+    /// Spanish side when the card is stored that way. Strip parenthetical
+    /// notes and the reflexive `-se` ending for verb-table lookup.
+    private func inferInfinitive(card: VocabCard) -> String {
+        let raw: String
+        if deck.isReversed {
+            raw = card.back
+        } else {
+            raw = card.front
+        }
+        var t = raw.trimmingCharacters(in: .whitespacesAndNewlines)
+        if let paren = t.firstIndex(of: "(") {
+            t = String(t[..<paren]).trimmingCharacters(in: .whitespacesAndNewlines)
+        }
+        if t.hasSuffix("se") && t.count > 4 { t = String(t.dropLast(2)) }
+        return t
+    }
+
     private func loadCards() {
         let deckId = deck.id
         let descriptor = FetchDescriptor<VocabCard>(
@@ -107,6 +131,36 @@ struct DeckStudyView: View {
                                     .multilineTextAlignment(.trailing)
                             }
 
+                            // Stem-change conjugation toggle
+                            if isStemChangingDeck {
+                                let verb = inferInfinitive(card: card)
+                                let isOpen = expandedConjugations.contains(verb)
+                                Button {
+                                    withAnimation(.smooth) {
+                                        if isOpen {
+                                            expandedConjugations.remove(verb)
+                                        } else {
+                                            expandedConjugations.insert(verb)
+                                        }
+                                    }
+                                } label: {
+                                    Label(
+                                        isOpen ? "Hide conjugation" : "Show conjugation",
+                                        systemImage: isOpen ? "chevron.up" : "chevron.down"
+                                    )
+                                    .font(.caption.weight(.medium))
+                                }
+                                .buttonStyle(.borderless)
+                                .tint(.blue)
+                                .padding(.leading, 42)
+
+                                if isOpen {
+                                    StemChangeConjugationView(infinitive: verb)
+                                        .padding(.leading, 42)
+                                        .transition(.opacity.combined(with: .move(edge: .top)))
+                                }
+                            }
+
                             // Example sentences
                             if !card.examplesES.isEmpty {
                                 VStack(alignment: .leading, spacing: 6) {

Conjuga/Conjuga/Views/Course/StemChangeConjugationView.swift

@@ -0,0 +1,97 @@
+import SwiftUI
+import SharedModels
+import SwiftData
+
+/// Shows the present-tense conjugation of a verb (identified by infinitive),
+/// with any irregular/stem-change spans highlighted. Designed to drop into
+/// stem-changing verb flashcards so learners can see the conjugation in-place.
+struct StemChangeConjugationView: View {
+    let infinitive: String
+
+    @Environment(\.modelContext) private var modelContext
+    @State private var rows: [ConjugationRow] = []
+
+    private static let personLabels = ["yo", "tú", "él/ella/Ud.", "nosotros", "vosotros", "ellos/ellas/Uds."]
+    private static let tenseId = "ind_presente"
+
+    var body: some View {
+        VStack(alignment: .leading, spacing: 8) {
+            HStack {
+                Text("Present tense")
+                    .font(.subheadline.weight(.semibold))
+                    .foregroundStyle(.secondary)
+                Spacer()
+            }
+            if rows.isEmpty {
+                Text("Conjugation not available")
+                    .font(.caption)
+                    .foregroundStyle(.secondary)
+                    .padding(.vertical, 4)
+            } else {
+                VStack(spacing: 6) {
+                    ForEach(rows) { row in
+                        HStack(alignment: .firstTextBaseline) {
+                            Text(row.person)
+                                .font(.callout)
+                                .foregroundStyle(.secondary)
+                                .frame(width: 130, alignment: .leading)
+                            IrregularHighlightText(
+                                form: row.form,
+                                spans: row.spans,
+                                font: .callout.monospaced(),
+                                showLabels: false
+                            )
+                            Spacer()
+                        }
+                    }
+                }
+            }
+        }
+        .padding(12)
+        .frame(maxWidth: .infinity, alignment: .leading)
+        .background(Color.blue.opacity(0.08), in: RoundedRectangle(cornerRadius: 10))
+        .onAppear(perform: loadForms)
+    }
+
+    private func loadForms() {
+        // Find the verb by infinitive (lowercase exact match).
+        let normalized = infinitive.lowercased().trimmingCharacters(in: .whitespaces)
+        let verbDescriptor = FetchDescriptor<Verb>(
+            predicate: #Predicate<Verb> { $0.infinitive == normalized }
+        )
+        guard let verb = (try? modelContext.fetch(verbDescriptor))?.first else {
+            rows = []
+            return
+        }
+
+        let verbId = verb.id
+        let tenseId = Self.tenseId
+        let formDescriptor = FetchDescriptor<VerbForm>(
+            predicate: #Predicate<VerbForm> { $0.verbId == verbId && $0.tenseId == tenseId },
+            sortBy: [SortDescriptor(\VerbForm.personIndex)]
+        )
+        let forms = (try? modelContext.fetch(formDescriptor)) ?? []
+
+        rows = forms.map { f in
+            ConjugationRow(
+                id: f.personIndex,
+                person: Self.personLabels[safe: f.personIndex] ?? "",
+                form: f.form,
+                spans: f.spans ?? []
+            )
+        }
+    }
+}
+
+private struct ConjugationRow: Identifiable {
+    let id: Int
+    let person: String
+    let form: String
+    let spans: [IrregularSpan]
+}
+
+private extension Array {
+    subscript(safe index: Int) -> Element? {
+        indices.contains(index) ? self[index] : nil
+    }
+}

Conjuga/Conjuga/Views/Course/TextbookChapterListView.swift

@@ -0,0 +1,121 @@
+import SwiftUI
+import SharedModels
+import SwiftData
+
+struct TextbookChapterListView: View {
+    let courseName: String
+
+    @Environment(\.cloudModelContextProvider) private var cloudModelContextProvider
+    @Query(sort: \TextbookChapter.number) private var allChapters: [TextbookChapter]
+
+    private var cloudModelContext: ModelContext { cloudModelContextProvider() }
+    @State private var attempts: [TextbookExerciseAttempt] = []
+
+    private var chapters: [TextbookChapter] {
+        allChapters.filter { $0.courseName == courseName }
+    }
+
+    private var byPart: [(part: Int, chapters: [TextbookChapter])] {
+        let grouped = Dictionary(grouping: chapters, by: \.part)
+        return grouped.keys.sorted().map { p in
+            (p, grouped[p]!.sorted { $0.number < $1.number })
+        }
+    }
+
+    private func progressFor(_ chapter: TextbookChapter) -> (correct: Int, total: Int) {
+        let chNum = chapter.number
+        let chAttempts = attempts.filter {
+            $0.courseName == courseName && $0.chapterNumber == chNum
+        }
+        let total = chAttempts.reduce(0) { $0 + $1.totalCount }
+        let correct = chAttempts.reduce(0) { $0 + $1.correctCount + $1.closeCount }
+        return (correct, total)
+    }
+
+    var body: some View {
+        List {
+            if chapters.isEmpty {
+                ContentUnavailableView(
+                    "Textbook loading",
+                    systemImage: "book.closed",
+                    description: Text("Textbook content is being prepared…")
+                )
+            } else {
+                ForEach(byPart, id: \.part) { part, partChapters in
+                    Section {
+                        ForEach(partChapters, id: \.id) { chapter in
+                            NavigationLink(value: chapter) {
+                                chapterRow(chapter)
+                            }
+                            .accessibilityIdentifier("textbook-chapter-row-\(chapter.number)")
+                        }
+                    } header: {
+                        if part > 0 {
+                            Text("Part \(part)")
+                        } else {
+                            Text("Chapters")
+                        }
+                    }
+                }
+            }
+        }
+        .navigationTitle("Textbook")
+        .onAppear(perform: loadAttempts)
+    }
+
+    @ViewBuilder
+    private func chapterRow(_ chapter: TextbookChapter) -> some View {
+        let p = progressFor(chapter)
+        HStack(alignment: .center, spacing: 12) {
+            ZStack {
+                Circle()
+                    .stroke(Color.secondary.opacity(0.2), lineWidth: 3)
+                    .frame(width: 36, height: 36)
+                if p.total > 0 {
+                    Circle()
+                        .trim(from: 0, to: CGFloat(p.correct) / CGFloat(p.total))
+                        .stroke(.orange, style: StrokeStyle(lineWidth: 3, lineCap: .round))
+                        .frame(width: 36, height: 36)
+                        .rotationEffect(.degrees(-90))
+                }
+                Text("\(chapter.number)")
+                    .font(.footnote.weight(.bold))
+            }
+
+            VStack(alignment: .leading, spacing: 2) {
+                Text(chapter.title)
+                    .font(.headline)
+                HStack(spacing: 10) {
+                    if chapter.exerciseCount > 0 {
+                        Label("\(chapter.exerciseCount)", systemImage: "pencil.and.list.clipboard")
+                            .font(.caption)
+                            .foregroundStyle(.secondary)
+                    }
+                    if chapter.vocabTableCount > 0 {
+                        Label("\(chapter.vocabTableCount)", systemImage: "list.bullet.rectangle")
+                            .font(.caption)
+                            .foregroundStyle(.secondary)
+                    }
+                    if p.total > 0 {
+                        Text("\(p.correct)/\(p.total)")
+                            .font(.caption.monospacedDigit())
+                            .foregroundStyle(.secondary)
+                    }
+                }
+            }
+            Spacer()
+        }
+        .padding(.vertical, 4)
+    }
+
+    private func loadAttempts() {
+        attempts = (try? cloudModelContext.fetch(FetchDescriptor<TextbookExerciseAttempt>())) ?? []
+    }
+}
+
+#Preview {
+    NavigationStack {
+        TextbookChapterListView(courseName: "Complete Spanish Step-by-Step")
+    }
+    .modelContainer(for: [TextbookChapter.self], inMemory: true)
+}

Conjuga/Conjuga/Views/Course/TextbookChapterView.swift

@@ -0,0 +1,209 @@
+import SwiftUI
+import SharedModels
+import SwiftData
+
+struct TextbookChapterView: View {
+    let chapter: TextbookChapter
+
+    @State private var expandedVocab: Set<Int> = []
+
+    private var blocks: [TextbookBlock] { chapter.blocks() }
+
+    var body: some View {
+        ScrollView {
+            VStack(alignment: .leading, spacing: 12) {
+                headerView
+                Divider()
+                ForEach(blocks) { block in
+                    blockView(block)
+                }
+            }
+            .padding(.horizontal)
+            .padding(.vertical, 12)
+        }
+        .navigationTitle(chapter.title)
+        .navigationBarTitleDisplayMode(.inline)
+    }
+
+    private var headerView: some View {
+        VStack(alignment: .leading, spacing: 4) {
+            if chapter.part > 0 {
+                Text("Part \(chapter.part)")
+                    .font(.subheadline)
+                    .foregroundStyle(.secondary)
+            }
+            Text("Chapter \(chapter.number)")
+                .font(.subheadline)
+                .foregroundStyle(.secondary)
+            Text(chapter.title)
+                .font(.largeTitle.bold())
+        }
+    }
+
+    @ViewBuilder
+    private func blockView(_ block: TextbookBlock) -> some View {
+        switch block.kind {
+        case .heading:
+            headingView(block)
+        case .paragraph:
+            paragraphView(block)
+        case .keyVocabHeader:
+            HStack(spacing: 6) {
+                Image(systemName: "star.fill").foregroundStyle(.orange)
+                Text("Key Vocabulary")
+                    .font(.headline)
+                    .foregroundStyle(.orange)
+            }
+            .padding(.top, 8)
+        case .vocabTable:
+            vocabTableView(block)
+        case .exercise:
+            exerciseLinkView(block)
+        }
+    }
+
+    private func headingView(_ block: TextbookBlock) -> some View {
+        let level = block.level ?? 3
+        let font: Font
+        switch level {
+        case 2: font = .title.bold()
+        case 3: font = .title2.bold()
+        case 4: font = .title3.weight(.semibold)
+        default: font = .headline
+        }
+        return Text(stripInlineEmphasis(block.text ?? ""))
+            .font(font)
+            .padding(.top, 10)
+    }
+
+    private func paragraphView(_ block: TextbookBlock) -> some View {
+        Text(attributedFromMarkdownish(block.text ?? ""))
+            .font(.body)
+            .fixedSize(horizontal: false, vertical: true)
+    }
+
+    private func vocabTableView(_ block: TextbookBlock) -> some View {
+        let expanded = expandedVocab.contains(block.index)
+        let cards = block.cards ?? []
+        let lines = block.ocrLines ?? []
+        let itemCount = cards.isEmpty ? lines.count : cards.count
+        return VStack(alignment: .leading, spacing: 4) {
+            Button {
+                if expanded { expandedVocab.remove(block.index) } else { expandedVocab.insert(block.index) }
+            } label: {
+                HStack {
+                    Image(systemName: expanded ? "chevron.down" : "chevron.right")
+                        .font(.caption)
+                    Text("Vocabulary (\(itemCount) items)")
+                        .font(.subheadline.weight(.medium))
+                        .foregroundStyle(.primary)
+                    Spacer()
+                }
+                .contentShape(Rectangle())
+            }
+            .buttonStyle(.plain)
+
+            if expanded {
+                if cards.isEmpty {
+                    // Fallback: no paired cards available — show raw OCR lines.
+                    VStack(alignment: .leading, spacing: 2) {
+                        ForEach(Array(lines.enumerated()), id: \.offset) { _, line in
+                            Text(line)
+                                .font(.callout.monospaced())
+                                .foregroundStyle(.secondary)
+                        }
+                    }
+                    .padding(.leading, 14)
+                } else {
+                    vocabGrid(cards: cards)
+                        .padding(.leading, 14)
+                }
+            }
+        }
+        .padding(10)
+        .frame(maxWidth: .infinity, alignment: .leading)
+        .background(Color.orange.opacity(0.08), in: RoundedRectangle(cornerRadius: 10))
+    }
+
+    @ViewBuilder
+    private func vocabGrid(cards: [TextbookVocabPair]) -> some View {
+        Grid(alignment: .leading, horizontalSpacing: 16, verticalSpacing: 6) {
+            ForEach(Array(cards.enumerated()), id: \.offset) { _, card in
+                GridRow {
+                    Text(card.front)
+                        .font(.callout)
+                        .foregroundStyle(.primary)
+                    Text(card.back)
+                        .font(.callout)
+                        .foregroundStyle(.secondary)
+                }
+            }
+        }
+    }
+
+    private func exerciseLinkView(_ block: TextbookBlock) -> some View {
+        NavigationLink(value: TextbookExerciseDestination(
+            chapterId: chapter.id,
+            chapterNumber: chapter.number,
+            blockIndex: block.index
+        )) {
+            HStack(spacing: 10) {
+                Image(systemName: "pencil.and.list.clipboard")
+                    .foregroundStyle(.orange)
+                    .font(.title3)
+                VStack(alignment: .leading, spacing: 2) {
+                    Text("Exercise \(block.exerciseId ?? "")")
+                        .font(.headline)
+                    if let inst = block.instruction, !inst.isEmpty {
+                        Text(stripInlineEmphasis(inst))
+                            .font(.caption)
+                            .foregroundStyle(.secondary)
+                            .lineLimit(2)
+                    }
+                }
+                Spacer()
+                Image(systemName: "chevron.right")
+                    .foregroundStyle(.secondary)
+                    .font(.caption)
+            }
+            .padding(12)
+            .background(Color.orange.opacity(0.1), in: RoundedRectangle(cornerRadius: 10))
+        }
+        .buttonStyle(.plain)
+    }
+
+    // Strip our ad-hoc ** / * markers from parsed text
+    private func stripInlineEmphasis(_ s: String) -> String {
+        s.replacingOccurrences(of: "**", with: "")
+            .replacingOccurrences(of: "*", with: "")
+    }
+
+    private func attributedFromMarkdownish(_ s: String) -> AttributedString {
+        // Parser emits `**bold**` and `*italic*`. Try to render via AttributedString markdown.
+        if let parsed = try? AttributedString(markdown: s, options: .init(allowsExtendedAttributes: true)) {
+            return parsed
+        }
+        return AttributedString(stripInlineEmphasis(s))
+    }
+}
+
+struct TextbookExerciseDestination: Hashable {
+    let chapterId: String
+    let chapterNumber: Int
+    let blockIndex: Int
+}
+
+#Preview {
+    NavigationStack {
+        TextbookChapterView(chapter: TextbookChapter(
+            id: "ch1",
+            number: 1,
+            title: "Sample",
+            part: 1,
+            courseName: "Preview",
+            bodyJSON: Data(),
+            exerciseCount: 0,
+            vocabTableCount: 0
+        ))
+    }
+}

Conjuga/Conjuga/Views/Course/TextbookExerciseView.swift

@@ -0,0 +1,360 @@
+import SwiftUI
+import SharedModels
+import SwiftData
+import PencilKit
+
+/// Interactive fill-in-the-blank view for one textbook exercise.
+/// Supports keyboard typing OR Apple Pencil handwriting input per prompt.
+struct TextbookExerciseView: View {
+    let chapter: TextbookChapter
+    let blockIndex: Int
+
+    @Environment(\.cloudModelContextProvider) private var cloudModelContextProvider
+    @State private var answers: [Int: String] = [:]
+    @State private var drawings: [Int: PKDrawing] = [:]
+    @State private var grades: [Int: TextbookGrade] = [:]
+    @State private var inputMode: InputMode = .keyboard
+    @State private var activePencilPromptNumber: Int?
+    @State private var isRecognizing = false
+    @State private var isChecked = false
+    @State private var recognizedTextForActive: String = ""
+
+    private var cloudModelContext: ModelContext { cloudModelContextProvider() }
+
+    enum InputMode: String {
+        case keyboard
+        case pencil
+    }
+
+    private var block: TextbookBlock? {
+        chapter.blocks().first { $0.index == blockIndex }
+    }
+
+    private var answerByNumber: [Int: TextbookAnswerItem] {
+        guard let items = block?.answerItems else { return [:] }
+        var out: [Int: TextbookAnswerItem] = [:]
+        for it in items {
+            out[it.number] = it
+        }
+        return out
+    }
+
+    var body: some View {
+        ScrollView {
+            VStack(alignment: .leading, spacing: 16) {
+                if let b = block {
+                    headerView(b)
+                    inputModePicker
+                    exerciseBody(b)
+                    checkButton(b)
+                } else {
+                    ContentUnavailableView(
+                        "Exercise not found",
+                        systemImage: "questionmark.circle"
+                    )
+                }
+            }
+            .padding()
+        }
+        .navigationTitle("Exercise \(block?.exerciseId ?? "")")
+        .navigationBarTitleDisplayMode(.inline)
+        .onAppear(perform: loadPreviousAttempt)
+    }
+
+    private func headerView(_ b: TextbookBlock) -> some View {
+        VStack(alignment: .leading, spacing: 8) {
+            Text("Chapter \(chapter.number): \(chapter.title)")
+                .font(.caption)
+                .foregroundStyle(.secondary)
+            Text("Exercise \(b.exerciseId ?? "")")
+                .font(.title2.bold())
+            if let inst = b.instruction, !inst.isEmpty {
+                Text(stripInlineEmphasis(inst))
+                    .font(.callout)
+                    .foregroundStyle(.secondary)
+                    .fixedSize(horizontal: false, vertical: true)
+            }
+            if let extra = b.extra, !extra.isEmpty {
+                ForEach(Array(extra.enumerated()), id: \.offset) { _, e in
+                    Text(stripInlineEmphasis(e))
+                        .font(.callout)
+                        .fixedSize(horizontal: false, vertical: true)
+                        .padding(8)
+                        .frame(maxWidth: .infinity, alignment: .leading)
+                        .background(Color.secondary.opacity(0.1), in: RoundedRectangle(cornerRadius: 8))
+                }
+            }
+        }
+    }
+
+    private var inputModePicker: some View {
+        Picker("Input mode", selection: $inputMode) {
+            Label("Keyboard", systemImage: "keyboard").tag(InputMode.keyboard)
+            Label("Pencil", systemImage: "pencil.tip").tag(InputMode.pencil)
+        }
+        .pickerStyle(.segmented)
+    }
+
+    private func exerciseBody(_ b: TextbookBlock) -> some View {
+        VStack(alignment: .leading, spacing: 14) {
+            if b.freeform == true {
+                VStack(alignment: .leading, spacing: 6) {
+                    Label("Freeform exercise", systemImage: "text.bubble")
+                        .font(.subheadline.weight(.semibold))
+                        .foregroundStyle(.orange)
+                    Text("Answers will vary. Use this space to write your own responses; they won't be auto-checked.")
+                        .font(.caption)
+                        .foregroundStyle(.secondary)
+                }
+                .padding()
+                .background(Color.orange.opacity(0.1), in: RoundedRectangle(cornerRadius: 10))
+            }
+            let rawPrompts = b.prompts ?? []
+            let prompts = rawPrompts.isEmpty ? synthesizedPrompts(b) : rawPrompts
+            if prompts.isEmpty && b.extra?.isEmpty == false {
+                Text("Fill in the blanks above; answers will be graded when you tap Check.")
+                    .font(.caption)
+                    .foregroundStyle(.secondary)
+            } else {
+                ForEach(Array(prompts.enumerated()), id: \.offset) { i, prompt in
+                    promptRow(index: i, prompt: prompt, expected: answerByNumber[i + 1])
+                }
+            }
+        }
+    }
+
+    /// When the source exercise prompts were embedded in a bitmap (common in
+    /// this textbook), we have no text for each question — only the answer
+    /// key. Synthesize numbered placeholders so the user still gets one input
+    /// field per answer.
+    private func synthesizedPrompts(_ b: TextbookBlock) -> [String] {
+        guard let items = b.answerItems, !items.isEmpty else { return [] }
+        return items.map { "\($0.number)." }
+    }
+
+    private func promptRow(index: Int, prompt: String, expected: TextbookAnswerItem?) -> some View {
+        let number = index + 1
+        let grade = grades[number]
+        return VStack(alignment: .leading, spacing: 8) {
+            HStack(alignment: .top, spacing: 8) {
+                if let grade {
+                    Image(systemName: iconFor(grade))
+                        .foregroundStyle(colorFor(grade))
+                        .font(.title3)
+                        .padding(.top, 2)
+                }
+                Text(stripInlineEmphasis(prompt))
+                    .font(.body)
+                    .fixedSize(horizontal: false, vertical: true)
+            }
+
+            switch inputMode {
+            case .keyboard:
+                TextField("Your answer", text: binding(for: number))
+                    .textFieldStyle(.roundedBorder)
+                    .textInputAutocapitalization(.never)
+                    .disableAutocorrection(true)
+                    .font(.body)
+                    .disabled(isChecked)
+            case .pencil:
+                pencilRow(number: number)
+            }
+
+            if isChecked, let grade, grade != .correct, let expected {
+                HStack(spacing: 6) {
+                    Text("Answer:")
+                        .font(.caption.weight(.semibold))
+                    Text(expected.answer)
+                        .font(.caption)
+                    if !expected.alternates.isEmpty {
+                        Text("(also: \(expected.alternates.joined(separator: ", ")))")
+                            .font(.caption2)
+                            .foregroundStyle(.secondary)
+                    }
+                }
+                .foregroundStyle(colorFor(grade))
+            }
+        }
+        .padding(10)
+        .background(backgroundFor(grade), in: RoundedRectangle(cornerRadius: 8))
+    }
+
+    private func pencilRow(number: Int) -> some View {
+        VStack(alignment: .leading, spacing: 6) {
+            HandwritingCanvas(
+                drawing: bindingDrawing(for: number),
+                onDrawingChanged: { recognizePencil(for: number) }
+            )
+            .frame(height: 100)
+            .background(.fill.quinary, in: RoundedRectangle(cornerRadius: 10))
+            .overlay(RoundedRectangle(cornerRadius: 10).stroke(.separator, lineWidth: 1))
+
+            HStack {
+                if let typed = answers[number], !typed.isEmpty {
+                    Text("Recognized: \(typed)")
+                        .font(.caption)
+                        .foregroundStyle(.secondary)
+                }
+                Spacer()
+                Button("Clear") {
+                    drawings[number] = PKDrawing()
+                    answers[number] = ""
+                }
+                .font(.caption)
+                .tint(.secondary)
+            }
+        }
+    }
+
+    private func checkButton(_ b: TextbookBlock) -> some View {
+        let hasAnyAnswer = answers.values.contains { !$0.isEmpty }
+        let disabled = b.freeform == true || (!isChecked && !hasAnyAnswer)
+        return Button {
+            if isChecked {
+                resetExercise()
+            } else {
+                checkAnswers(b)
+            }
+        } label: {
+            Text(isChecked ? "Try again" : "Check answers")
+                .font(.headline)
+                .frame(maxWidth: .infinity)
+                .padding(.vertical, 10)
+        }
+        .buttonStyle(.borderedProminent)
+        .tint(.orange)
+        .disabled(disabled)
+    }
+
+    // MARK: - Actions
+
+    private func checkAnswers(_ b: TextbookBlock) {
+        guard let prompts = b.prompts else { return }
+        var newGrades: [Int: TextbookGrade] = [:]
+        var states: [TextbookPromptState] = []
+        for (i, _) in prompts.enumerated() {
+            let number = i + 1
+            let user = answers[number] ?? ""
+            let expected = answerByNumber[number]
+            let canonical = expected?.answer ?? ""
+            let alts = expected?.alternates ?? []
+            let grade: TextbookGrade
+            if canonical.isEmpty {
+                grade = .wrong
+            } else {
+                grade = AnswerChecker.grade(userText: user, canonical: canonical, alternates: alts)
+            }
+            newGrades[number] = grade
+            states.append(TextbookPromptState(number: number, userText: user, grade: grade))
+        }
+        grades = newGrades
+        isChecked = true
+        saveAttempt(states: states, exerciseId: b.exerciseId ?? "")
+    }
+
+    private func resetExercise() {
+        answers.removeAll()
+        drawings.removeAll()
+        grades.removeAll()
+        isChecked = false
+    }
+
+    private func recognizePencil(for number: Int) {
+        guard let drawing = drawings[number], !drawing.strokes.isEmpty else { return }
+        isRecognizing = true
+        Task {
+            let result = await HandwritingRecognizer.recognize(drawing: drawing)
+            await MainActor.run {
+                answers[number] = result.text
+                isRecognizing = false
+            }
+        }
+    }
+
+    private func saveAttempt(states: [TextbookPromptState], exerciseId: String) {
+        let attemptId = TextbookExerciseAttempt.attemptId(
+            courseName: chapter.courseName,
+            exerciseId: exerciseId
+        )
+        let descriptor = FetchDescriptor<TextbookExerciseAttempt>(
+            predicate: #Predicate<TextbookExerciseAttempt> { $0.id == attemptId }
+        )
+        let context = cloudModelContext
+        let existing = (try? context.fetch(descriptor))?.first
+        let attempt = existing ?? TextbookExerciseAttempt(
+            id: attemptId,
+            courseName: chapter.courseName,
+            chapterNumber: chapter.number,
+            exerciseId: exerciseId
+        )
+        if existing == nil { context.insert(attempt) }
+        attempt.lastAttemptAt = Date()
+        attempt.setPromptStates(states)
+        try? context.save()
+    }
+
+    private func loadPreviousAttempt() {
+        guard let b = block else { return }
+        let attemptId = TextbookExerciseAttempt.attemptId(
+            courseName: chapter.courseName,
+            exerciseId: b.exerciseId ?? ""
+        )
+        let descriptor = FetchDescriptor<TextbookExerciseAttempt>(
+            predicate: #Predicate<TextbookExerciseAttempt> { $0.id == attemptId }
+        )
+        guard let attempt = (try? cloudModelContext.fetch(descriptor))?.first else { return }
+        for s in attempt.promptStates() {
+            answers[s.number] = s.userText
+            grades[s.number] = s.grade
+        }
+        isChecked = !grades.isEmpty
+    }
+
+    // MARK: - Bindings
+
+    private func binding(for number: Int) -> Binding<String> {
+        Binding(
+            get: { answers[number] ?? "" },
+            set: { answers[number] = $0 }
+        )
+    }
+
+    private func bindingDrawing(for number: Int) -> Binding<PKDrawing> {
+        Binding(
+            get: { drawings[number] ?? PKDrawing() },
+            set: { drawings[number] = $0 }
+        )
+    }
+
+    // MARK: - UI helpers
+
+    private func iconFor(_ grade: TextbookGrade) -> String {
+        switch grade {
+        case .correct: return "checkmark.circle.fill"
+        case .close: return "circle.lefthalf.filled"
+        case .wrong: return "xmark.circle.fill"
+        }
+    }
+
+    private func colorFor(_ grade: TextbookGrade) -> Color {
+        switch grade {
+        case .correct: return .green
+        case .close: return .orange
+        case .wrong: return .red
+        }
+    }
+
+    private func backgroundFor(_ grade: TextbookGrade?) -> Color {
+        guard let grade else { return Color.secondary.opacity(0.05) }
+        switch grade {
+        case .correct: return .green.opacity(0.12)
+        case .close: return .orange.opacity(0.12)
+        case .wrong: return .red.opacity(0.12)
+        }
+    }
+
+    private func stripInlineEmphasis(_ s: String) -> String {
+        s.replacingOccurrences(of: "**", with: "")
+            .replacingOccurrences(of: "*", with: "")
+    }
+}

Conjuga/Conjuga/Views/Course/VocabFlashcardView.swift

@@ -6,11 +6,19 @@ struct VocabFlashcardView: View {
     let cards: [VocabCard]
     let speechService: SpeechService
     let onDone: () -> Void
+    /// Optional deck context — when present and the title indicates a stem-
+    /// changing deck, each card gets an inline conjugation toggle.
+    var deckTitle: String? = nil
 
     @Environment(\.cloudModelContextProvider) private var cloudModelContextProvider
     @State private var currentIndex = 0
     @State private var isRevealed = false
     @State private var sessionCorrect = 0
+    @State private var showConjugation = false
+
+    private var isStemChangingDeck: Bool {
+        (deckTitle ?? "").localizedCaseInsensitiveContains("stem changing")
+    }
 
     private var cloudModelContext: ModelContext { cloudModelContextProvider() }
 
@@ -61,6 +69,25 @@ struct VocabFlashcardView: View {
                                         .padding(12)
                                 }
                                 .glassEffect(in: .circle)
+
+                                if isStemChangingDeck {
+                                    Button {
+                                        withAnimation(.smooth) { showConjugation.toggle() }
+                                    } label: {
+                                        Label(
+                                            showConjugation ? "Hide conjugation" : "Show conjugation",
+                                            systemImage: showConjugation ? "chevron.up" : "chevron.down"
+                                        )
+                                        .font(.subheadline.weight(.medium))
+                                    }
+                                    .buttonStyle(.bordered)
+                                    .tint(.blue)
+
+                                    if showConjugation {
+                                        StemChangeConjugationView(infinitive: stripToInfinitive(card.front))
+                                            .transition(.opacity.combined(with: .move(edge: .top)))
+                                    }
+                                }
                             }
                             .transition(.blurReplace)
                         } else {
@@ -111,6 +138,7 @@ struct VocabFlashcardView: View {
                                     guard currentIndex > 0 else { return }
                                     withAnimation(.smooth) {
                                         isRevealed = false
+                                        showConjugation = false
                                         currentIndex -= 1
                                     }
                                 } label: {
@@ -125,6 +153,7 @@ struct VocabFlashcardView: View {
                                 Button {
                                     withAnimation(.smooth) {
                                         isRevealed = false
+                                        showConjugation = false
                                         currentIndex += 1
                                     }
                                 } label: {
@@ -189,9 +218,25 @@ struct VocabFlashcardView: View {
         // Next card
         withAnimation(.smooth) {
             isRevealed = false
+            showConjugation = false
             currentIndex += 1
         }
     }
+
+    /// Card fronts may be plain infinitives ("cerrar") or, in reversed decks,
+    /// stored as English. Strip any reflexive-se suffix or parenthetical notes
+    /// to improve the verb lookup hit rate.
+    private func stripToInfinitive(_ s: String) -> String {
+        var t = s.trimmingCharacters(in: .whitespacesAndNewlines)
+        if let paren = t.firstIndex(of: "(") {
+            t = String(t[..<paren]).trimmingCharacters(in: .whitespacesAndNewlines)
+        }
+        if t.hasSuffix("se") && t.count > 4 {
+            // "acostarse" → "acostar" for verb lookup
+            t = String(t.dropLast(2))
+        }
+        return t
+    }
 }
 
 #Preview {

Conjuga/Conjuga/Views/Onboarding/OnboardingView.swift

@@ -128,7 +128,7 @@ struct OnboardingView: View {
 
     private func completeOnboarding() {
         let progress = ReviewStore.fetchOrCreateUserProgress(context: cloudModelContext)
-        progress.selectedVerbLevel = selectedLevel
+        progress.selectedVerbLevels = [selectedLevel]
         if progress.enabledTenseIDs.isEmpty {
             progress.enabledTenseIDs = ReviewStore.defaultEnabledTenses()
         }

Conjuga/Conjuga/Views/Practice/Chat/ChatView.swift

@@ -1,16 +1,27 @@
 import SwiftUI
 import SharedModels
 import SwiftData
+import FoundationModels
+
+@Generable
+private struct ChatWordInfo {
+    @Guide(description: "Dictionary base form") var baseForm: String
+    @Guide(description: "English translation") var english: String
+    @Guide(description: "Part of speech") var partOfSpeech: String
+}
 
 struct ChatView: View {
     let conversation: Conversation
 
     @Environment(\.cloudModelContextProvider) private var cloudModelContextProvider
+    @Environment(DictionaryService.self) private var dictionary
     @State private var service = ConversationService()
     @State private var messages: [ChatMessage] = []
     @State private var inputText = ""
     @State private var errorMessage: String?
     @State private var hasStarted = false
+    @State private var selectedWord: WordAnnotation?
+    @State private var lookupCache: [String: WordAnnotation] = [:]
 
     private var cloudContext: ModelContext { cloudModelContextProvider() }
 
@@ -21,7 +32,9 @@ struct ChatView: View {
                 ScrollView {
                     LazyVStack(spacing: 12) {
                         ForEach(messages) { message in
-                            ChatBubble(message: message)
+                            ChatBubble(message: message, dictionary: dictionary, lookupCache: $lookupCache) { word in
+                                selectedWord = word
+                            }
                             .id(message.id)
                         }
 
@@ -68,6 +81,10 @@ struct ChatView: View {
         }
         .navigationTitle(conversation.scenario)
         .navigationBarTitleDisplayMode(.inline)
+        .sheet(item: $selectedWord) { word in
+            ChatWordDetailSheet(word: word)
+                .presentationDetents([.height(200)])
+        }
         .alert("Error", isPresented: .init(
             get: { errorMessage != nil },
             set: { if !$0 { errorMessage = nil } }
@@ -121,6 +138,9 @@ struct ChatView: View {
 
 private struct ChatBubble: View {
     let message: ChatMessage
+    let dictionary: DictionaryService
+    @Binding var lookupCache: [String: WordAnnotation]
+    let onWordTap: (WordAnnotation) -> Void
 
     private var isUser: Bool { message.role == "user" }
 
@@ -129,11 +149,15 @@ private struct ChatBubble: View {
             if isUser { Spacer(minLength: 60) }
 
             VStack(alignment: isUser ? .trailing : .leading, spacing: 4) {
+                if isUser {
                     Text(message.content)
                         .font(.body)
                         .padding(.horizontal, 14)
                         .padding(.vertical, 10)
-                    .background(isUser ? AnyShapeStyle(.green.opacity(0.2)) : AnyShapeStyle(.fill.quaternary), in: RoundedRectangle(cornerRadius: 16))
+                        .background(.green.opacity(0.2), in: RoundedRectangle(cornerRadius: 16))
+                } else {
+                    tappableBubble
+                }
 
                 if let correction = message.correction, !correction.isEmpty {
                     Text(correction)
@@ -147,4 +171,179 @@ private struct ChatBubble: View {
         }
         .padding(.horizontal)
     }
+
+    private var tappableBubble: some View {
+        let words = message.content.components(separatedBy: " ")
+        return ChatFlowLayout(spacing: 0) {
+            ForEach(Array(words.enumerated()), id: \.offset) { _, word in
+                ChatWordButton(word: word, dictionary: dictionary, cache: lookupCache) { annotation in
+                    if annotation.english.isEmpty {
+                        lookupWordAsync(annotation.word)
+                    } else {
+                        onWordTap(annotation)
+                    }
+                }
+            }
+        }
+        .padding(.horizontal, 14)
+        .padding(.vertical, 10)
+        .background(.fill.quaternary, in: RoundedRectangle(cornerRadius: 16))
+    }
+
+    private func lookupWordAsync(_ word: String) {
+        // Try dictionary first
+        if let entry = dictionary.lookup(word) {
+            let annotation = WordAnnotation(word: word, baseForm: entry.baseForm, english: entry.english, partOfSpeech: entry.partOfSpeech)
+            lookupCache[word] = annotation
+            onWordTap(annotation)
+            return
+        }
+
+        // Show loading then AI lookup
+        onWordTap(WordAnnotation(word: word, baseForm: word, english: "Looking up...", partOfSpeech: ""))
+
+        Task {
+            do {
+                let session = LanguageModelSession(instructions: "You are a Spanish dictionary. Provide base form, English translation, and part of speech.")
+                let response = try await session.respond(to: "Word: \"\(word)\"", generating: ChatWordInfo.self)
+                let info = response.content
+                let annotation = WordAnnotation(word: word, baseForm: info.baseForm, english: info.english, partOfSpeech: info.partOfSpeech)
+                lookupCache[word] = annotation
+                onWordTap(annotation)
+            } catch {
+                onWordTap(WordAnnotation(word: word, baseForm: word, english: "Lookup unavailable", partOfSpeech: ""))
+            }
+        }
+    }
+}
+
+// MARK: - Chat Word Button
+
+private struct ChatWordButton: View {
+    let word: String
+    let dictionary: DictionaryService
+    let cache: [String: WordAnnotation]
+    let onTap: (WordAnnotation) -> Void
+
+    private var cleaned: String {
+        word.lowercased()
+            .trimmingCharacters(in: .punctuationCharacters)
+            .trimmingCharacters(in: .whitespaces)
+    }
+
+    private var annotation: WordAnnotation? {
+        if let cached = cache[cleaned] { return cached }
+        if let entry = dictionary.lookup(cleaned) {
+            return WordAnnotation(word: cleaned, baseForm: entry.baseForm, english: entry.english, partOfSpeech: entry.partOfSpeech)
+        }
+        return nil
+    }
+
+    var body: some View {
+        Button {
+            onTap(annotation ?? WordAnnotation(word: cleaned, baseForm: cleaned, english: "", partOfSpeech: ""))
+        } label: {
+            Text(word + " ")
+                .font(.body)
+                .foregroundStyle(.primary)
+                .underline(annotation != nil, color: .teal.opacity(0.3))
+        }
+        .buttonStyle(.plain)
+    }
+}
+
+// MARK: - Word Detail Sheet
+
+private struct ChatWordDetailSheet: View {
+    let word: WordAnnotation
+
+    var body: some View {
+        VStack(spacing: 16) {
+            HStack {
+                Text(word.word)
+                    .font(.title2.bold())
+                Spacer()
+                if !word.partOfSpeech.isEmpty {
+                    Text(word.partOfSpeech)
+                        .font(.caption.weight(.medium))
+                        .foregroundStyle(.secondary)
+                        .padding(.horizontal, 8)
+                        .padding(.vertical, 4)
+                        .background(.fill.tertiary, in: Capsule())
+                }
+            }
+
+            Divider()
+
+            if word.english == "Looking up..." {
+                HStack(spacing: 8) {
+                    ProgressView()
+                    Text("Looking up word...")
+                        .font(.subheadline)
+                        .foregroundStyle(.secondary)
+                }
+                .frame(maxWidth: .infinity)
+            } else {
+                VStack(alignment: .leading, spacing: 8) {
+                    if !word.baseForm.isEmpty && word.baseForm != word.word {
+                        HStack {
+                            Text("Base form:")
+                                .font(.subheadline)
+                                .foregroundStyle(.secondary)
+                            Text(word.baseForm)
+                                .font(.subheadline.weight(.semibold))
+                                .italic()
+                        }
+                    }
+                    if !word.english.isEmpty {
+                        HStack {
+                            Text("English:")
+                                .font(.subheadline)
+                                .foregroundStyle(.secondary)
+                            Text(word.english)
+                                .font(.subheadline.weight(.semibold))
+                        }
+                    }
+                }
+                .frame(maxWidth: .infinity, alignment: .leading)
+            }
+
+            Spacer()
+        }
+        .padding()
+    }
+}
+
+// MARK: - Chat Flow Layout
+
+private struct ChatFlowLayout: Layout {
+    var spacing: CGFloat = 0
+    func sizeThatFits(proposal: ProposedViewSize, subviews: Subviews, cache: inout ()) -> CGSize {
+        let rows = computeRows(proposal: proposal, subviews: subviews)
+        var height: CGFloat = 0
+        for row in rows { height += row.map { $0.height }.max() ?? 0 }
+        height += CGFloat(max(0, rows.count - 1)) * spacing
+        return CGSize(width: proposal.width ?? 0, height: height)
+    }
+    func placeSubviews(in bounds: CGRect, proposal: ProposedViewSize, subviews: Subviews, cache: inout ()) {
+        let rows = computeRows(proposal: proposal, subviews: subviews)
+        var y = bounds.minY; var idx = 0
+        for row in rows {
+            var x = bounds.minX; let rh = row.map { $0.height }.max() ?? 0
+            for size in row {
+                subviews[idx].place(at: CGPoint(x: x, y: y), proposal: ProposedViewSize(size))
+                x += size.width; idx += 1
+            }
+            y += rh + spacing
+        }
+    }
+    private func computeRows(proposal: ProposedViewSize, subviews: Subviews) -> [[CGSize]] {
+        let mw = proposal.width ?? .infinity; var rows: [[CGSize]] = [[]]; var cw: CGFloat = 0
+        for sv in subviews {
+            let s = sv.sizeThatFits(.unspecified)
+            if cw + s.width > mw && !rows[rows.count - 1].isEmpty { rows.append([]); cw = 0 }
+            rows[rows.count - 1].append(s); cw += s.width
+        }
+        return rows
+    }
 }

Conjuga/Conjuga/Views/Practice/FullTableView.swift

@@ -243,7 +243,11 @@ struct FullTableView: View {
         results = Array(repeating: nil, count: 6)
         correctForms = []
         drawings = Array(repeating: PKDrawing(), count: 6)
-        let service = PracticeSessionService(localContext: modelContext, cloudContext: cloudModelContext)
+        let service = PracticeSessionService(
+            localContext: modelContext,
+            cloudContext: cloudModelContext,
+            reflexiveBaseInfinitives: ReflexiveVerbStore.shared.baseInfinitives
+        )
         guard let prompt = service.randomFullTablePrompt() else {
             currentVerb = nil
             currentTense = nil
@@ -312,7 +316,11 @@ struct FullTableView: View {
         if allCorrect { sessionCorrect += 1 }
 
         if let verb = currentVerb, let tense = currentTense {
-            let service = PracticeSessionService(localContext: modelContext, cloudContext: cloudModelContext)
+            let service = PracticeSessionService(
+                localContext: modelContext,
+                cloudContext: cloudModelContext,
+                reflexiveBaseInfinitives: ReflexiveVerbStore.shared.baseInfinitives
+            )
             let reviewResults = Dictionary(uniqueKeysWithValues: personsToShow.map { ($0.index, results[$0.index] == true) })
             _ = service.recordFullTableReview(verbId: verb.id, tenseId: tense.id, results: reviewResults)
         }

Conjuga/Conjuga/Views/Practice/Lyrics/LyricsReaderView.swift

@@ -4,6 +4,10 @@ import SharedModels
 struct LyricsReaderView: View {
     let song: SavedSong
 
+    @Environment(DictionaryService.self) private var dictionary
+    @State private var selectedWord: LyricsWordLookup?
+    @State private var lookupCache: [String: LyricsWordLookup] = [:]
+
     var body: some View {
         ScrollView {
             VStack(spacing: 20) {
@@ -15,6 +19,10 @@ struct LyricsReaderView: View {
         }
         .navigationTitle(song.title)
         .navigationBarTitleDisplayMode(.inline)
+        .sheet(item: $selectedWord) { word in
+            LyricsWordDetailSheet(word: word)
+                .presentationDetents([.height(260)])
+        }
     }
 
     // MARK: - Header
@@ -56,15 +64,6 @@ struct LyricsReaderView: View {
         let spanishLines = song.lyricsES.components(separatedBy: "\n")
         let englishLines = song.lyricsEN.components(separatedBy: "\n")
         let lineCount = max(spanishLines.count, englishLines.count)
-        let _ = {
-            print("[LyricsReader] ES lines: \(spanishLines.count), EN lines: \(englishLines.count), rendering: \(lineCount)")
-            for i in 0..<min(15, lineCount) {
-                let es = i < spanishLines.count ? spanishLines[i] : "(none)"
-                let en = i < englishLines.count ? englishLines[i] : "(none)"
-                print("  [\(i)] ES: \(es.isEmpty ? "(blank)" : es)")
-                print("       EN: \(en.isEmpty ? "(blank)" : en)")
-            }
-        }()
 
         return VStack(alignment: .leading, spacing: 0) {
             ForEach(0..<lineCount, id: \.self) { index in
@@ -78,8 +77,7 @@ struct LyricsReaderView: View {
                 } else {
                     VStack(alignment: .leading, spacing: 2) {
                         if !es.isEmpty {
-                            Text(es)
+                            spanishLine(es)
-                                .font(.body.weight(.medium))
                         }
                         if !en.isEmpty {
                             Text(en)
@@ -94,4 +92,183 @@ struct LyricsReaderView: View {
         .padding()
         .background(.fill.quinary, in: RoundedRectangle(cornerRadius: 12))
     }
+
+    private func spanishLine(_ line: String) -> some View {
+        let tokens = line.components(separatedBy: " ")
+        return LyricsFlowLayout(spacing: 0) {
+            ForEach(Array(tokens.enumerated()), id: \.offset) { _, token in
+                LyricsWordView(token: token, lookup: makeLookup(for: token)) { word in
+                    selectedWord = word
+                }
+            }
+        }
+    }
+
+    // MARK: - Lookup
+
+    private func makeLookup(for rawToken: String) -> LyricsWordLookup? {
+        let cleaned = rawToken.lowercased()
+            .trimmingCharacters(in: .punctuationCharacters)
+            .trimmingCharacters(in: .whitespaces)
+        guard !cleaned.isEmpty else { return nil }
+
+        if let cached = lookupCache[cleaned] { return cached }
+        guard let entry = dictionary.lookup(cleaned) else { return nil }
+
+        let displayWord = rawToken
+            .trimmingCharacters(in: .punctuationCharacters)
+            .trimmingCharacters(in: .whitespaces)
+
+        let tenseDisplay = entry.tenseId.flatMap { TenseInfo.find($0)?.english }
+
+        let lookup = LyricsWordLookup(
+            word: displayWord.isEmpty ? entry.word : displayWord,
+            baseForm: entry.baseForm,
+            english: entry.english,
+            partOfSpeech: entry.partOfSpeech,
+            tenseDisplay: tenseDisplay,
+            person: entry.person
+        )
+        lookupCache[cleaned] = lookup
+        return lookup
+    }
+}
+
+// MARK: - Word Lookup Model
+
+private struct LyricsWordLookup: Identifiable, Hashable {
+    let word: String
+    let baseForm: String
+    let english: String
+    let partOfSpeech: String
+    let tenseDisplay: String?
+    let person: String?
+
+    var id: String { word }
+}
+
+// MARK: - Word View
+
+private struct LyricsWordView: View {
+    let token: String
+    let lookup: LyricsWordLookup?
+    let onLookup: (LyricsWordLookup) -> Void
+
+    var body: some View {
+        Text(token + " ")
+            .font(.body.weight(.medium))
+            .foregroundStyle(.primary)
+            .underline(lookup != nil, color: .teal.opacity(0.35))
+            .contentShape(Rectangle())
+            .onLongPressGesture(minimumDuration: 0.35) {
+                if let lookup {
+                    onLookup(lookup)
+                }
+            }
+    }
+}
+
+// MARK: - Detail Sheet
+
+private struct LyricsWordDetailSheet: View {
+    let word: LyricsWordLookup
+
+    var body: some View {
+        VStack(alignment: .leading, spacing: 14) {
+            HStack {
+                Text(word.word)
+                    .font(.title2.bold())
+                Spacer()
+                if !word.partOfSpeech.isEmpty {
+                    Text(word.partOfSpeech)
+                        .font(.caption.weight(.medium))
+                        .foregroundStyle(.secondary)
+                        .padding(.horizontal, 8)
+                        .padding(.vertical, 4)
+                        .background(.fill.tertiary, in: Capsule())
+                }
+            }
+
+            Divider()
+
+            VStack(alignment: .leading, spacing: 10) {
+                if !word.baseForm.isEmpty && word.baseForm.lowercased() != word.word.lowercased() {
+                    detailRow(label: "Base form", value: word.baseForm, italic: true)
+                }
+
+                if !word.english.isEmpty {
+                    detailRow(label: "English", value: word.english)
+                }
+
+                if let tenseDisplay = word.tenseDisplay {
+                    let personSuffix = (word.person?.isEmpty == false) ? " · \(word.person!)" : ""
+                    detailRow(label: "Tense", value: tenseDisplay + personSuffix)
+                }
+            }
+
+            Spacer(minLength: 0)
+        }
+        .padding()
+        .frame(maxWidth: .infinity, alignment: .leading)
+    }
+
+    @ViewBuilder
+    private func detailRow(label: String, value: String, italic: Bool = false) -> some View {
+        HStack(alignment: .firstTextBaseline, spacing: 8) {
+            Text("\(label):")
+                .font(.subheadline)
+                .foregroundStyle(.secondary)
+                .frame(width: 86, alignment: .leading)
+            Text(value)
+                .font(.subheadline.weight(.semibold))
+                .italic(italic)
+            Spacer(minLength: 0)
+        }
+    }
+}
+
+// MARK: - Flow Layout
+
+private struct LyricsFlowLayout: Layout {
+    var spacing: CGFloat = 0
+
+    func sizeThatFits(proposal: ProposedViewSize, subviews: Subviews, cache: inout ()) -> CGSize {
+        let rows = computeRows(proposal: proposal, subviews: subviews)
+        var height: CGFloat = 0
+        for row in rows { height += row.map { $0.height }.max() ?? 0 }
+        height += CGFloat(max(0, rows.count - 1)) * spacing
+        return CGSize(width: proposal.width ?? 0, height: height)
+    }
+
+    func placeSubviews(in bounds: CGRect, proposal: ProposedViewSize, subviews: Subviews, cache: inout ()) {
+        let rows = computeRows(proposal: proposal, subviews: subviews)
+        var y = bounds.minY
+        var idx = 0
+        for row in rows {
+            var x = bounds.minX
+            let rh = row.map { $0.height }.max() ?? 0
+            for size in row {
+                subviews[idx].place(at: CGPoint(x: x, y: y), proposal: ProposedViewSize(size))
+                x += size.width
+                idx += 1
+            }
+            y += rh + spacing
+        }
+    }
+
+    private func computeRows(proposal: ProposedViewSize, subviews: Subviews) -> [[CGSize]] {
+        let mw = proposal.width ?? .infinity
+        var rows: [[CGSize]] = [[]]
+        var cw: CGFloat = 0
+        for sv in subviews {
+            let s = sv.sizeThatFits(.unspecified)
+            if cw + s.width > mw && !rows[rows.count - 1].isEmpty {
+                rows.append([])
+                cw = 0
+            }
+            rows[rows.count - 1].append(s)
+            cw += s.width
+        }
+        return rows
+    }
 }

Conjuga/Conjuga/Views/Settings/SettingsView.swift

@@ -9,9 +9,11 @@ struct SettingsView: View {
     @State private var dailyGoal: Double = 50
     @State private var showVosotros: Bool = true
     @State private var autoFillStem: Bool = false
-    @State private var selectedLevel: VerbLevel = .basic
 
     private let levels = VerbLevel.allCases
+    private let irregularCategories: [IrregularSpan.SpanCategory] = [
+        .spelling, .stemChange, .uniqueIrregular
+    ]
     private var cloudModelContext: ModelContext { cloudModelContextProvider() }
 
     var body: some View {
@@ -40,19 +42,26 @@ struct SettingsView: View {
                         }
                 }
 
-                Section("Level") {
+                Section {
-                    Picker("Current Level", selection: $selectedLevel) {
                     ForEach(levels, id: \.self) { level in
-                            Text(level.displayName).tag(level)
+                        Toggle(level.displayName, isOn: Binding(
-                        }
+                            get: {
-                    }
+                                progress?.selectedVerbLevels.contains(level) ?? false
-                    .onChange(of: selectedLevel) { _, newValue in
+                            },
-                        progress?.selectedVerbLevel = newValue
+                            set: { enabled in
+                                guard let progress else { return }
+                                progress.setLevelEnabled(level, enabled: enabled)
                                 saveProgress()
                             }
+                        ))
+                    }
+                } header: {
+                    Text("Levels")
+                } footer: {
+                    Text("Practice pulls only from verbs whose level is enabled. Turn on multiple to mix.")
                 }
 
-                Section("Tenses") {
+                Section {
                     ForEach(TenseInfo.all) { tense in
                         Toggle(tense.english, isOn: Binding(
                             get: {
@@ -65,6 +74,41 @@ struct SettingsView: View {
                             }
                         ))
                     }
+                } header: {
+                    Text("Tenses")
+                }
+
+                Section {
+                    ForEach(irregularCategories, id: \.self) { category in
+                        Toggle(category.rawValue, isOn: Binding(
+                            get: {
+                                progress?.enabledIrregularCategories.contains(category) ?? false
+                            },
+                            set: { enabled in
+                                guard let progress else { return }
+                                progress.setIrregularCategoryEnabled(category, enabled: enabled)
+                                saveProgress()
+                            }
+                        ))
+                    }
+                } header: {
+                    Text("Irregular Types")
+                } footer: {
+                    Text("Leave all off to include regular and irregular verbs. Enable any to restrict practice to those irregularity types.")
+                }
+
+                Section {
+                    Toggle("Reflexive verbs only", isOn: Binding(
+                        get: { progress?.showReflexiveVerbsOnly ?? false },
+                        set: { enabled in
+                            progress?.showReflexiveVerbsOnly = enabled
+                            saveProgress()
+                        }
+                    ))
+                } header: {
+                    Text("Reflexive")
+                } footer: {
+                    Text("When on, practice pulls only from the curated list of common reflexive verbs.")
                 }
 
                 Section("Stats") {
@@ -96,7 +140,6 @@ struct SettingsView: View {
         dailyGoal = Double(resolved.dailyGoal)
         showVosotros = resolved.showVosotros
         autoFillStem = resolved.autoFillStem
-        selectedLevel = resolved.selectedVerbLevel
     }
 
     private func saveProgress() {

Conjuga/Conjuga/Views/Verbs/VerbDetailView.swift

@@ -4,14 +4,40 @@ import SwiftData
 
 struct VerbDetailView: View {
     @Environment(\.modelContext) private var modelContext
+    @Environment(VerbExampleCache.self) private var exampleCache
+    @Environment(ReflexiveVerbStore.self) private var reflexiveStore
     @State private var speechService = SpeechService()
     let verb: Verb
     @State private var selectedTense: TenseInfo = TenseInfo.all[0]
 
+    @State private var examples: [VerbExample] = []
+    @State private var examplesState: ExamplesState = .idle
+
+    private enum ExamplesState: Equatable {
+        case idle
+        case loading
+        case loaded
+        case unavailable
+        case failed(String)
+    }
+
+    private static let exampleTenseIds: [String] = [
+        TenseID.ind_presente.rawValue,
+        TenseID.ind_preterito.rawValue,
+        TenseID.ind_imperfecto.rawValue,
+        TenseID.ind_futuro.rawValue,
+        TenseID.subj_presente.rawValue,
+        TenseID.imp_afirmativo.rawValue,
+    ]
+
     private var formsForTense: [VerbForm] {
         ReferenceStore(context: modelContext).fetchForms(verbId: verb.id, tenseId: selectedTense.id)
     }
 
+    private var reflexiveEntries: [ReflexiveVerb] {
+        reflexiveStore.entries(for: verb.infinitive)
+    }
+
     var body: some View {
         List {
             Section {
@@ -25,6 +51,10 @@ struct VerbDetailView: View {
                 Text("Info")
             }
 
+            if !reflexiveEntries.isEmpty {
+                reflexiveSection
+            }
+
             Section {
                 Picker("Tense", selection: $selectedTense) {
                     ForEach(TenseInfo.all) { tense in
@@ -66,6 +96,8 @@ struct VerbDetailView: View {
             } header: {
                 Text("Conjugation")
             }
+
+            examplesSection
         }
         .navigationTitle(verb.infinitive)
         .toolbar {
@@ -78,6 +110,126 @@ struct VerbDetailView: View {
                 .tint(.secondary)
             }
         }
+        .task(id: verb.id) {
+            await loadExamples()
+        }
+    }
+
+    // MARK: - Reflexive
+
+    private var reflexiveSection: some View {
+        Section {
+            ForEach(Array(reflexiveEntries.enumerated()), id: \.offset) { _, entry in
+                VStack(alignment: .leading, spacing: 4) {
+                    HStack(alignment: .firstTextBaseline, spacing: 8) {
+                        Text(entry.infinitive)
+                            .font(.body.weight(.semibold))
+                            .italic()
+                        if let hint = entry.usageHint, !hint.isEmpty {
+                            Text(hint)
+                                .font(.caption.weight(.medium))
+                                .foregroundStyle(.tint)
+                        }
+                    }
+                    Text(entry.english)
+                        .font(.caption)
+                        .foregroundStyle(.secondary)
+                }
+                .padding(.vertical, 2)
+            }
+        } header: {
+            Text("Reflexive")
+        } footer: {
+            if reflexiveEntries.contains(where: { $0.usageHint != nil }) {
+                Text("Highlighted words are prepositions or phrases this verb commonly pairs with.")
+                    .font(.caption2)
+            }
+        }
+    }
+
+    // MARK: - Examples
+
+    @ViewBuilder
+    private var examplesSection: some View {
+        Section {
+            switch examplesState {
+            case .idle, .loading:
+                HStack(spacing: 10) {
+                    ProgressView()
+                    Text("Generating examples…")
+                        .font(.caption)
+                        .foregroundStyle(.secondary)
+                }
+            case .unavailable:
+                Label("Examples require Apple Intelligence on this device.", systemImage: "sparkles")
+                    .font(.caption)
+                    .foregroundStyle(.secondary)
+            case .failed(let message):
+                Label(message, systemImage: "exclamationmark.triangle")
+                    .font(.caption)
+                    .foregroundStyle(.secondary)
+            case .loaded:
+                if examples.isEmpty {
+                    Label("No examples available.", systemImage: "text.quote")
+                        .font(.caption)
+                        .foregroundStyle(.secondary)
+                } else {
+                    ForEach(Array(examples.enumerated()), id: \.offset) { _, example in
+                        VStack(alignment: .leading, spacing: 4) {
+                            if let info = TenseInfo.find(example.tenseId) {
+                                Text(info.english)
+                                    .font(.caption2.weight(.semibold))
+                                    .foregroundStyle(.tint)
+                            }
+                            Text(example.spanish)
+                                .font(.body)
+                                .italic()
+                            Text(example.english)
+                                .font(.caption)
+                                .foregroundStyle(.secondary)
+                        }
+                        .padding(.vertical, 2)
+                    }
+                }
+            }
+        } header: {
+            Text("Examples")
+        }
+    }
+
+    private func loadExamples() async {
+        // Reset state when navigating between verbs via NavigationSplitView.
+        examples = []
+        examplesState = .idle
+
+        if let cached = exampleCache.examples(for: verb.id), !cached.isEmpty {
+            examples = cached
+            examplesState = .loaded
+            return
+        }
+
+        guard VerbExampleGenerator.isAvailable else {
+            examplesState = .unavailable
+            return
+        }
+
+        examplesState = .loading
+        do {
+            let generated = try await VerbExampleGenerator.generate(
+                verbInfinitive: verb.infinitive,
+                verbEnglish: verb.english,
+                tenseIds: Self.exampleTenseIds
+            )
+            guard !generated.isEmpty else {
+                examplesState = .failed("Could not generate examples.")
+                return
+            }
+            exampleCache.setExamples(generated, for: verb.id)
+            examples = generated
+            examplesState = .loaded
+        } catch {
+            examplesState = .failed("Could not generate examples.")
+        }
     }
 }
 
@@ -86,4 +238,6 @@ struct VerbDetailView: View {
         VerbDetailView(verb: Verb(id: 1, infinitive: "hablar", english: "to speak", rank: 1, ending: "ar", reflexive: 0, level: "basic"))
     }
     .modelContainer(for: [Verb.self, VerbForm.self], inMemory: true)
+    .environment(VerbExampleCache())
+    .environment(ReflexiveVerbStore())
 }

Conjuga/Conjuga/Views/Verbs/VerbListView.swift

@@ -2,11 +2,33 @@ import SwiftUI
 import SwiftData
 import SharedModels
 
+enum IrregularityCategory: String, CaseIterable, Identifiable {
+    case anyIrregular = "Any Irregular"
+    case spelling = "Spelling Change"
+    case stemChange = "Stem Change"
+    case uniqueIrregular = "Unique Irregular"
+
+    var id: String { rawValue }
+
+    var systemImage: String {
+        switch self {
+        case .anyIrregular: "asterisk"
+        case .spelling: "character.cursor.ibeam"
+        case .stemChange: "arrow.triangle.2.circlepath"
+        case .uniqueIrregular: "star"
+        }
+    }
+}
+
 struct VerbListView: View {
     @Environment(\.modelContext) private var modelContext
+    @Environment(ReflexiveVerbStore.self) private var reflexiveStore
     @State private var verbs: [Verb] = []
+    @State private var irregularityByVerbId: [Int: Set<IrregularityCategory>] = [:]
     @State private var searchText = ""
     @State private var selectedLevel: String?
+    @State private var selectedIrregularity: IrregularityCategory?
+    @State private var reflexiveOnly: Bool = false
     @State private var selectedVerb: Verb?
 
     private var filteredVerbs: [Verb] {
@@ -14,6 +36,15 @@ struct VerbListView: View {
         if let level = selectedLevel {
             result = result.filter { VerbLevelGroup.matches($0.level, selectedLevel: level) }
         }
+        if let category = selectedIrregularity {
+            result = result.filter { verb in
+                guard let cats = irregularityByVerbId[verb.id] else { return false }
+                return category == .anyIrregular ? !cats.isEmpty : cats.contains(category)
+            }
+        }
+        if reflexiveOnly {
+            result = result.filter { reflexiveStore.isReflexive(baseInfinitive: $0.infinitive) }
+        }
         if !searchText.isEmpty {
             let query = searchText.lowercased()
             result = result.filter {
@@ -30,20 +61,58 @@ struct VerbListView: View {
         NavigationSplitView {
             List(filteredVerbs, selection: $selectedVerb) { verb in
                 NavigationLink(value: verb) {
-                    VerbRowView(verb: verb)
+                    VerbRowView(verb: verb, irregularities: irregularityByVerbId[verb.id] ?? [])
                 }
             }
             .navigationTitle("Verbs")
             .searchable(text: $searchText, prompt: "Search verbs...")
+            .safeAreaInset(edge: .top, spacing: 0) {
+                if hasActiveFilter {
+                    activeFilterBar
+                }
+            }
             .toolbar {
                 ToolbarItem(placement: .topBarTrailing) {
                     Menu {
-                        Button("All Levels") { selectedLevel = nil }
+                        Section("Level") {
+                            Button {
+                                selectedLevel = nil
+                            } label: {
+                                Label("All Levels", systemImage: selectedLevel == nil ? "checkmark" : "")
+                            }
                             ForEach(levels, id: \.self) { level in
-                            Button(level.capitalized) { selectedLevel = level }
+                                Button {
+                                    selectedLevel = level
+                                } label: {
+                                    Label(level.capitalized, systemImage: selectedLevel == level ? "checkmark" : "")
+                                }
+                            }
+                        }
+
+                        Section("Irregularity") {
+                            Button {
+                                selectedIrregularity = nil
+                            } label: {
+                                Label("All Verbs", systemImage: selectedIrregularity == nil ? "checkmark" : "")
+                            }
+                            ForEach(IrregularityCategory.allCases) { category in
+                                Button {
+                                    selectedIrregularity = category
+                                } label: {
+                                    Label(category.rawValue, systemImage: selectedIrregularity == category ? "checkmark" : category.systemImage)
+                                }
+                            }
+                        }
+
+                        Section("Reflexive") {
+                            Button {
+                                reflexiveOnly.toggle()
+                            } label: {
+                                Label("Reflexive verbs only", systemImage: reflexiveOnly ? "checkmark" : "arrow.triangle.2.circlepath")
+                            }
                         }
                     } label: {
-                        Label("Filter", systemImage: "line.3.horizontal.decrease.circle")
+                        Label("Filter", systemImage: hasActiveFilter ? "line.3.horizontal.decrease.circle.fill" : "line.3.horizontal.decrease.circle")
                     }
                 }
             }
@@ -58,6 +127,56 @@ struct VerbListView: View {
         }
     }
 
+    private var hasActiveFilter: Bool {
+        selectedLevel != nil || selectedIrregularity != nil || reflexiveOnly
+    }
+
+    @ViewBuilder
+    private var activeFilterBar: some View {
+        HStack(spacing: 8) {
+            if let level = selectedLevel {
+                filterChip(text: level.capitalized, systemImage: "graduationcap") {
+                    selectedLevel = nil
+                }
+            }
+            if let cat = selectedIrregularity {
+                filterChip(text: cat.rawValue, systemImage: cat.systemImage) {
+                    selectedIrregularity = nil
+                }
+            }
+            if reflexiveOnly {
+                filterChip(text: "Reflexive", systemImage: "arrow.triangle.2.circlepath") {
+                    reflexiveOnly = false
+                }
+            }
+            Spacer()
+            Text("\(filteredVerbs.count)")
+                .font(.caption.monospacedDigit())
+                .foregroundStyle(.secondary)
+        }
+        .padding(.horizontal)
+        .padding(.vertical, 8)
+        .background(.bar)
+    }
+
+    private func filterChip(text: String, systemImage: String, onClear: @escaping () -> Void) -> some View {
+        Button(action: onClear) {
+            HStack(spacing: 4) {
+                Image(systemName: systemImage)
+                    .font(.caption2)
+                Text(text)
+                    .font(.caption.weight(.medium))
+                Image(systemName: "xmark")
+                    .font(.caption2)
+            }
+            .padding(.horizontal, 10)
+            .padding(.vertical, 5)
+            .background(.blue.opacity(0.15), in: Capsule())
+            .foregroundStyle(.blue)
+        }
+        .buttonStyle(.plain)
+    }
+
     private func loadVerbs() {
         // Hit the shared local container directly, bypassing @Environment.
         guard let container = SharedStore.localContainer else {
@@ -69,12 +188,30 @@ struct VerbListView: View {
         }
         let context = ModelContext(container)
         verbs = ReferenceStore(context: context).fetchVerbs()
-        print("[VerbListView] loaded \(verbs.count) verbs (container: \(ObjectIdentifier(container)))")
+        irregularityByVerbId = buildIrregularityIndex(context: context)
+        print("[VerbListView] loaded \(verbs.count) verbs, \(irregularityByVerbId.count) flagged irregular (container: \(ObjectIdentifier(container)))")
+    }
+
+    private func buildIrregularityIndex(context: ModelContext) -> [Int: Set<IrregularityCategory>] {
+        let spans = (try? context.fetch(FetchDescriptor<IrregularSpan>())) ?? []
+        var index: [Int: Set<IrregularityCategory>] = [:]
+        for span in spans {
+            let category: IrregularityCategory
+            switch span.spanType {
+            case 100..<200: category = .spelling
+            case 200..<300: category = .stemChange
+            case 300..<400: category = .uniqueIrregular
+            default: continue
+            }
+            index[span.verbId, default: []].insert(category)
+        }
+        return index
     }
 }
 
 struct VerbRowView: View {
     let verb: Verb
+    var irregularities: Set<IrregularityCategory> = []
 
     var body: some View {
         HStack {
@@ -88,6 +225,15 @@ struct VerbRowView: View {
 
             Spacer()
 
+            HStack(spacing: 4) {
+                ForEach(orderedIrregularities, id: \.self) { cat in
+                    Image(systemName: cat.systemImage)
+                        .font(.caption2.weight(.semibold))
+                        .foregroundStyle(irregularityColor(cat))
+                        .help(cat.rawValue)
+                        .accessibilityLabel(cat.rawValue)
+                }
+
                 Text(verb.level.prefix(3).uppercased())
                     .font(.caption2)
                     .fontWeight(.semibold)
@@ -98,6 +244,22 @@ struct VerbRowView: View {
                     .clipShape(Capsule())
             }
         }
+    }
+
+    private var orderedIrregularities: [IrregularityCategory] {
+        // Order: unique > stem > spelling (most notable first)
+        let order: [IrregularityCategory] = [.uniqueIrregular, .stemChange, .spelling]
+        return order.filter { irregularities.contains($0) }
+    }
+
+    private func irregularityColor(_ category: IrregularityCategory) -> Color {
+        switch category {
+        case .uniqueIrregular: return .purple
+        case .stemChange: return .orange
+        case .spelling: return .teal
+        case .anyIrregular: return .gray
+        }
+    }
 
     private func levelColor(_ level: String) -> Color {
         switch level {

Conjuga/Conjuga/reflexive_verbs.json

@@ -0,0 +1,104 @@
+[
+  {"infinitive": "aburrirse", "baseInfinitive": "aburrir", "english": "to get bored"},
+  {"infinitive": "acercarse", "baseInfinitive": "acercar", "english": "to get close to", "usageHint": "a"},
+  {"infinitive": "acordarse", "baseInfinitive": "acordar", "english": "to remember", "usageHint": "de"},
+  {"infinitive": "acostarse", "baseInfinitive": "acostar", "english": "to lay down / to go to bed"},
+  {"infinitive": "acostumbrarse", "baseInfinitive": "acostumbrar", "english": "to get used to", "usageHint": "a"},
+  {"infinitive": "afeitarse", "baseInfinitive": "afeitar", "english": "to shave"},
+  {"infinitive": "alegrarse", "baseInfinitive": "alegrar", "english": "to be glad / happy / pleased"},
+  {"infinitive": "alejarse", "baseInfinitive": "alejar", "english": "to get away from", "usageHint": "de"},
+  {"infinitive": "animarse", "baseInfinitive": "animar", "english": "to cheer up / to dare to do something", "usageHint": "a"},
+  {"infinitive": "apurarse", "baseInfinitive": "apurar", "english": "to hurry"},
+  {"infinitive": "aprovecharse", "baseInfinitive": "aprovechar", "english": "to take advantage of", "usageHint": "de"},
+  {"infinitive": "asustarse", "baseInfinitive": "asustar", "english": "to get or become afraid"},
+  {"infinitive": "atreverse", "baseInfinitive": "atrever", "english": "to dare to", "usageHint": "a"},
+  {"infinitive": "bañarse", "baseInfinitive": "bañar", "english": "to take a bath / shower"},
+  {"infinitive": "burlarse", "baseInfinitive": "burlar", "english": "to make fun of", "usageHint": "de"},
+  {"infinitive": "caerse", "baseInfinitive": "caer", "english": "to fall down"},
+  {"infinitive": "calmarse", "baseInfinitive": "calmar", "english": "to calm down"},
+  {"infinitive": "cansarse", "baseInfinitive": "cansar", "english": "to get tired (of)", "usageHint": "(de)"},
+  {"infinitive": "casarse", "baseInfinitive": "casar", "english": "to marry", "usageHint": "con"},
+  {"infinitive": "cepillarse", "baseInfinitive": "cepillar", "english": "to brush (hair, teeth)"},
+  {"infinitive": "deprimirse", "baseInfinitive": "deprimir", "english": "to become depressed"},
+  {"infinitive": "conformarse", "baseInfinitive": "conformar", "english": "to resign oneself to", "usageHint": "con"},
+  {"infinitive": "volverse", "baseInfinitive": "volver", "english": "to become / to turn into / to return"},
+  {"infinitive": "darse", "baseInfinitive": "dar", "english": "to realize", "usageHint": "cuenta de"},
+  {"infinitive": "dedicarse", "baseInfinitive": "dedicar", "english": "to dedicate oneself to / to do for a living", "usageHint": "a"},
+  {"infinitive": "despedirse", "baseInfinitive": "despedir", "english": "to say goodbye", "usageHint": "(de)"},
+  {"infinitive": "despertarse", "baseInfinitive": "despertar", "english": "to wake up"},
+  {"infinitive": "desvestirse", "baseInfinitive": "desvestir", "english": "to undress"},
+  {"infinitive": "dirigirse", "baseInfinitive": "dirigir", "english": "to go to / make one's way toward / to address", "usageHint": "a"},
+  {"infinitive": "hacerse", "baseInfinitive": "hacer", "english": "to become / to pretend"},
+  {"infinitive": "divertirse", "baseInfinitive": "divertir", "english": "to have fun"},
+  {"infinitive": "dormirse", "baseInfinitive": "dormir", "english": "to fall asleep / to oversleep"},
+  {"infinitive": "ducharse", "baseInfinitive": "duchar", "english": "to shower"},
+  {"infinitive": "echarse", "baseInfinitive": "echar", "english": "to begin (usually suddenly) to do something / to break into", "usageHint": "a"},
+  {"infinitive": "enamorarse", "baseInfinitive": "enamorar", "english": "to fall in love with", "usageHint": "de"},
+  {"infinitive": "encargarse", "baseInfinitive": "encargar", "english": "to take charge of or be responsible for", "usageHint": "de"},
+  {"infinitive": "encogerse", "baseInfinitive": "encoger", "english": "to shrug (shoulders)", "usageHint": "(de hombros)"},
+  {"infinitive": "encontrarse", "baseInfinitive": "encontrar", "english": "to meet with / to run into someone", "usageHint": "(con)"},
+  {"infinitive": "enfermarse", "baseInfinitive": "enfermar", "english": "to get sick"},
+  {"infinitive": "enojarse", "baseInfinitive": "enojar", "english": "to get or become angry"},
+  {"infinitive": "enterarse", "baseInfinitive": "enterar", "english": "to find out, to realize", "usageHint": "de"},
+  {"infinitive": "exponerse", "baseInfinitive": "exponer", "english": "to expose oneself to or run the risk of", "usageHint": "a"},
+  {"infinitive": "fijarse", "baseInfinitive": "fijar", "english": "to pay attention to / to take a look"},
+  {"infinitive": "jugarse", "baseInfinitive": "jugar", "english": "to risk"},
+  {"infinitive": "lastimarse", "baseInfinitive": "lastimar", "english": "to get hurt or hurt oneself"},
+  {"infinitive": "lavarse", "baseInfinitive": "lavar", "english": "to wash (a body part)"},
+  {"infinitive": "levantarse", "baseInfinitive": "levantar", "english": "to get up"},
+  {"infinitive": "maquillarse", "baseInfinitive": "maquillar", "english": "to put makeup on"},
+  {"infinitive": "meterse", "baseInfinitive": "meter", "english": "to get into / to pick on / to pick a fight with", "usageHint": "en / con"},
+  {"infinitive": "motivarse", "baseInfinitive": "motivar", "english": "to become or get motivated to"},
+  {"infinitive": "moverse", "baseInfinitive": "mover", "english": "to move oneself"},
+  {"infinitive": "mudarse", "baseInfinitive": "mudar", "english": "to move (change residence)"},
+  {"infinitive": "negarse", "baseInfinitive": "negar", "english": "to refuse to", "usageHint": "a"},
+  {"infinitive": "obsesionarse", "baseInfinitive": "obsesionar", "english": "to be or get obsessed with", "usageHint": "con"},
+  {"infinitive": "ocuparse", "baseInfinitive": "ocupar", "english": "to look after", "usageHint": "de"},
+  {"infinitive": "olvidarse", "baseInfinitive": "olvidar", "english": "to forget", "usageHint": "de"},
+  {"infinitive": "parecerse", "baseInfinitive": "parecer", "english": "to look like someone or something", "usageHint": "a"},
+  {"infinitive": "peinarse", "baseInfinitive": "peinar", "english": "to comb your hair"},
+  {"infinitive": "ponerse", "baseInfinitive": "poner", "english": "to put on (clothing) / to get or become"},
+  {"infinitive": "ponerse", "baseInfinitive": "poner", "english": "to come to an agreement with someone", "usageHint": "de acuerdo"},
+  {"infinitive": "preocuparse", "baseInfinitive": "preocupar", "english": "to worry about", "usageHint": "por"},
+  {"infinitive": "prepararse", "baseInfinitive": "preparar", "english": "to prepare to"},
+  {"infinitive": "probarse", "baseInfinitive": "probar", "english": "to try on"},
+  {"infinitive": "quebrarse", "baseInfinitive": "quebrar", "english": "to break (an arm, leg, etc.)"},
+  {"infinitive": "quejarse", "baseInfinitive": "quejar", "english": "to complain about", "usageHint": "de"},
+  {"infinitive": "quedarse", "baseInfinitive": "quedar", "english": "to remain / to stay"},
+  {"infinitive": "quemarse", "baseInfinitive": "quemar", "english": "to burn oneself / one's body"},
+  {"infinitive": "quitarse", "baseInfinitive": "quitar", "english": "to take off (clothing, etc.)"},
+  {"infinitive": "reírse", "baseInfinitive": "reír", "english": "to laugh about", "usageHint": "de"},
+  {"infinitive": "resignarse", "baseInfinitive": "resignar", "english": "to resign oneself to", "usageHint": "a"},
+  {"infinitive": "romperse", "baseInfinitive": "romper", "english": "to break (an arm, leg, etc.)"},
+  {"infinitive": "secarse", "baseInfinitive": "secar", "english": "to dry (a body part)"},
+  {"infinitive": "sentarse", "baseInfinitive": "sentar", "english": "to sit down"},
+  {"infinitive": "sentirse", "baseInfinitive": "sentir", "english": "to feel"},
+  {"infinitive": "servirse", "baseInfinitive": "servir", "english": "to help oneself to (food)"},
+  {"infinitive": "suicidarse", "baseInfinitive": "suicidar", "english": "to commit suicide"},
+  {"infinitive": "tratarse", "baseInfinitive": "tratar", "english": "to be about", "usageHint": "de"},
+  {"infinitive": "vestirse", "baseInfinitive": "vestir", "english": "to get dressed"},
+  {"infinitive": "marearse", "baseInfinitive": "marear", "english": "to get sick, to get dizzy"},
+  {"infinitive": "irse", "baseInfinitive": "ir", "english": "to leave"},
+  {"infinitive": "imaginarse", "baseInfinitive": "imaginar", "english": "to imagine"},
+  {"infinitive": "preguntarse", "baseInfinitive": "preguntar", "english": "to wonder"},
+  {"infinitive": "llamarse", "baseInfinitive": "llamar", "english": "to be called"},
+  {"infinitive": "verse", "baseInfinitive": "ver", "english": "to look or appear"},
+  {"infinitive": "distraerse", "baseInfinitive": "distraer", "english": "to get distracted"},
+  {"infinitive": "concentrarse", "baseInfinitive": "concentrar", "english": "to focus"},
+  {"infinitive": "rendirse", "baseInfinitive": "rendir", "english": "to give up"},
+  {"infinitive": "relajarse", "baseInfinitive": "relajar", "english": "to relax"},
+  {"infinitive": "merecerse", "baseInfinitive": "merecer", "english": "to deserve"},
+  {"infinitive": "suponerse", "baseInfinitive": "suponer", "english": "to suppose"},
+  {"infinitive": "conectarse", "baseInfinitive": "conectar", "english": "to connect"},
+  {"infinitive": "destacarse", "baseInfinitive": "destacar", "english": "to stand out"},
+  {"infinitive": "recibirse", "baseInfinitive": "recibir", "english": "to graduate"},
+  {"infinitive": "graduarse", "baseInfinitive": "graduar", "english": "to graduate"},
+  {"infinitive": "perderse", "baseInfinitive": "perder", "english": "to get lost"},
+  {"infinitive": "cambiarse", "baseInfinitive": "cambiar", "english": "to change (clothing)", "usageHint": "(de ropa)"},
+  {"infinitive": "adaptarse", "baseInfinitive": "adaptar", "english": "to adapt, to adjust", "usageHint": "a"},
+  {"infinitive": "salirse", "baseInfinitive": "salir", "english": "to get away with", "usageHint": "con (la suya)"},
+  {"infinitive": "subirse", "baseInfinitive": "subir", "english": "to get on (the bus, etc.)", "usageHint": "a"},
+  {"infinitive": "tranquilizarse", "baseInfinitive": "tranquilizar", "english": "to relax"},
+  {"infinitive": "equivocarse", "baseInfinitive": "equivocar", "english": "to get something wrong / confused"},
+  {"infinitive": "confundirse", "baseInfinitive": "confundir", "english": "to get something wrong / confused"}
+]

Conjuga/ConjugaUITests/AllChaptersScreenshotTests.swift

@@ -0,0 +1,95 @@
+import XCTest
+
+/// Screenshot every chapter of the textbook — one top + one bottom frame each —
+/// so you can visually audit parsing / rendering issues across all 30 chapters.
+final class AllChaptersScreenshotTests: XCTestCase {
+
+    override func setUpWithError() throws {
+        continueAfterFailure = true
+    }
+
+    func testScreenshotEveryChapter() throws {
+        let app = XCUIApplication()
+        app.launchArguments += ["-onboardingComplete", "YES"]
+        app.launch()
+
+        let courseTab = app.tabBars.buttons["Course"]
+        XCTAssertTrue(courseTab.waitForExistence(timeout: 5))
+        courseTab.tap()
+
+        let textbookRow = app.buttons.containing(NSPredicate(
+            format: "label CONTAINS[c] 'Complete Spanish'"
+        )).firstMatch
+        XCTAssertTrue(textbookRow.waitForExistence(timeout: 5))
+        textbookRow.tap()
+
+        // NOTE: SwiftUI List preserves scroll position across navigation pushes,
+        // so visiting chapters in-order means the next one is already visible
+        // after we return from the previous one. No need to reset.
+        attach(app, name: "00-chapter-list-top")
+
+        for chapter in 1...30 {
+            guard let row = findChapterRow(app: app, chapter: chapter) else {
+                XCTFail("Chapter \(chapter) row not reachable")
+                continue
+            }
+            row.tap()
+
+            // Chapter body — wait until the chapter's title appears as a nav bar label
+            _ = app.navigationBars.firstMatch.waitForExistence(timeout: 3)
+
+            attach(app, name: String(format: "ch%02d-top", chapter))
+            // One big scroll to sample the bottom of the chapter
+            dragFullScreen(app, direction: .up)
+            dragFullScreen(app, direction: .up)
+            attach(app, name: String(format: "ch%02d-bottom", chapter))
+
+            tapNavBack(app)
+            // Small settle wait
+            _ = app.navigationBars.firstMatch.waitForExistence(timeout: 2)
+        }
+    }
+
+    // MARK: - Helpers
+
+    private enum DragDirection { case up, down }
+
+    private func dragFullScreen(_ app: XCUIApplication, direction: DragDirection) {
+        let top = app.coordinate(withNormalizedOffset: CGVector(dx: 0.5, dy: 0.12))
+        let bot = app.coordinate(withNormalizedOffset: CGVector(dx: 0.5, dy: 0.88))
+        switch direction {
+        case .up:   bot.press(forDuration: 0.1, thenDragTo: top)
+        case .down: top.press(forDuration: 0.1, thenDragTo: bot)
+        }
+    }
+
+    private func findChapterRow(app: XCUIApplication, chapter: Int) -> XCUIElement? {
+        // Chapter row accessibility label: "<n>, <title>, ..." (SwiftUI composes
+        // label from inner Texts). Match by starting number.
+        let predicate = NSPredicate(format: "label BEGINSWITH %@", "\(chapter),")
+        let row = app.buttons.matching(predicate).firstMatch
+
+        if row.exists && row.isHittable { return row }
+
+        // Scroll down up to 8 times searching for the row — chapters visited
+        // in order, so usually 0–2 swipes suffice.
+        for _ in 0..<8 {
+            if row.exists && row.isHittable { return row }
+            dragFullScreen(app, direction: .up)
+        }
+        return row.exists ? row : nil
+    }
+
+    private func tapNavBack(_ app: XCUIApplication) {
+        let back = app.navigationBars.buttons.firstMatch
+        if back.exists && back.isHittable { back.tap() }
+    }
+
+    private func attach(_ app: XCUIApplication, name: String) {
+        let screenshot = app.screenshot()
+        let attachment = XCTAttachment(screenshot: screenshot)
+        attachment.name = name
+        attachment.lifetime = .keepAlways
+        add(attachment)
+    }
+}

Conjuga/ConjugaUITests/StemChangeToggleTests.swift

@@ -0,0 +1,66 @@
+import XCTest
+
+final class StemChangeToggleTests: XCTestCase {
+
+    override func setUpWithError() throws {
+        continueAfterFailure = false
+    }
+
+    func testStemChangeConjugationToggle() throws {
+        let app = XCUIApplication()
+        app.launchArguments += ["-onboardingComplete", "YES"]
+        app.launch()
+
+        // Course → LanGo Beginner I → Week 4 → E-IE stem-changing verbs
+        app.tabBars.buttons["Course"].tap()
+
+        // Locate the E-IE deck row. Deck titles appear as static text / button.
+        // Scroll until visible, then tap.
+        let deckPredicate = NSPredicate(format: "label CONTAINS[c] 'E-IE stem changing verbs' AND NOT label CONTAINS[c] 'REVÉS'")
+        let deckRow = app.buttons.matching(deckPredicate).firstMatch
+
+        let listRef = app.coordinate(withNormalizedOffset: CGVector(dx: 0.5, dy: 0.85))
+        let topRef = app.coordinate(withNormalizedOffset: CGVector(dx: 0.5, dy: 0.10))
+        for _ in 0..<12 {
+            if deckRow.exists && deckRow.isHittable { break }
+            listRef.press(forDuration: 0.1, thenDragTo: topRef)
+        }
+        XCTAssertTrue(deckRow.waitForExistence(timeout: 3), "E-IE deck row missing")
+        deckRow.tap()
+
+        attach(app, name: "01-deck-top")
+
+        // Tap "Show conjugation" on the first card
+        let showBtn = app.buttons.matching(NSPredicate(format: "label BEGINSWITH 'Show conjugation'")).firstMatch
+        XCTAssertTrue(showBtn.waitForExistence(timeout: 3), "Show conjugation button missing")
+        showBtn.tap()
+
+        // Wait for the conjugation rows + animation to settle.
+        let yoLabel = app.staticTexts["yo"].firstMatch
+        XCTAssertTrue(yoLabel.waitForExistence(timeout: 3), "yo row not rendered")
+        // Give the transition time to complete before snapshotting.
+        Thread.sleep(forTimeInterval: 0.6)
+        attach(app, name: "02-conjugation-open")
+
+        // Also confirm all expected person labels are rendered.
+        for person in ["yo", "tú", "nosotros"] {
+            XCTAssertTrue(
+                app.staticTexts[person].firstMatch.exists,
+                "missing conjugation row for \(person)"
+            )
+        }
+
+        // Tap again to hide
+        let hideBtn = app.buttons.matching(NSPredicate(format: "label BEGINSWITH 'Hide conjugation'")).firstMatch
+        XCTAssertTrue(hideBtn.waitForExistence(timeout: 2))
+        hideBtn.tap()
+    }
+
+    private func attach(_ app: XCUIApplication, name: String) {
+        let s = app.screenshot()
+        let a = XCTAttachment(screenshot: s)
+        a.name = name
+        a.lifetime = .keepAlways
+        add(a)
+    }
+}

Conjuga/ConjugaUITests/TextbookFlowUITests.swift

@@ -0,0 +1,80 @@
+import XCTest
+
+final class TextbookFlowUITests: XCTestCase {
+
+    override func setUpWithError() throws {
+        continueAfterFailure = false
+    }
+
+    func testTextbookFlow() throws {
+        let app = XCUIApplication()
+        // Skip onboarding via defaults (already set by run script, but harmless to override)
+        app.launchArguments += ["-onboardingComplete", "YES"]
+        app.launch()
+
+        // Dashboard should be default tab. Switch to Course.
+        let courseTab = app.tabBars.buttons["Course"]
+        XCTAssertTrue(courseTab.waitForExistence(timeout: 5), "Course tab missing")
+        courseTab.tap()
+
+        // Attach a screenshot of the Course list
+        attach(app, name: "01-course-list")
+
+        // Tap the Textbook entry
+        let textbookRow = app.buttons.containing(NSPredicate(
+            format: "label CONTAINS[c] 'Complete Spanish'"
+        )).firstMatch
+        XCTAssertTrue(textbookRow.waitForExistence(timeout: 5), "Textbook row missing in Course")
+        textbookRow.tap()
+
+        attach(app, name: "02-textbook-chapter-list")
+
+        // Tap chapter 1 — should navigate to reader
+        let chapterOneRow = app.buttons.containing(NSPredicate(
+            format: "label CONTAINS[c] 'Nouns, Articles'"
+        )).firstMatch
+        XCTAssertTrue(chapterOneRow.waitForExistence(timeout: 5), "Chapter 1 row missing")
+        chapterOneRow.tap()
+
+        attach(app, name: "03-chapter-body")
+
+        // Find the first exercise link ("Exercise 1.1")
+        let exerciseRow = app.buttons.containing(NSPredicate(
+            format: "label CONTAINS[c] 'Exercise 1.1'"
+        )).firstMatch
+        XCTAssertTrue(exerciseRow.waitForExistence(timeout: 5), "Exercise 1.1 link missing")
+        exerciseRow.tap()
+
+        attach(app, name: "04-exercise-view")
+
+        // Check presence of input fields: at least a few numbered prompts
+        // Text fields use SwiftUI placeholder "Your answer"
+        let firstField = app.textFields["Your answer"].firstMatch
+        XCTAssertTrue(firstField.waitForExistence(timeout: 5), "No input fields rendered for exercise")
+        firstField.tap()
+        firstField.typeText("el")
+
+        attach(app, name: "05-exercise-typed-el")
+
+        // Tap Check answers
+        let checkButton = app.buttons["Check answers"]
+        XCTAssertTrue(checkButton.waitForExistence(timeout: 3), "Check answers button missing")
+        checkButton.tap()
+
+        attach(app, name: "06-exercise-graded")
+
+        // The first answer to Exercise 1.1 is "el" — we should see the first prompt
+        // graded correct. Iterating too deeply is fragile; just take a screenshot
+        // and check for presence of either a checkmark-like label or "Try again".
+        let tryAgain = app.buttons["Try again"]
+        XCTAssertTrue(tryAgain.waitForExistence(timeout: 3), "Grading did not complete")
+    }
+
+    private func attach(_ app: XCUIApplication, name: String) {
+        let screenshot = app.screenshot()
+        let attachment = XCTAttachment(screenshot: screenshot)
+        attachment.name = name
+        attachment.lifetime = .keepAlways
+        add(attachment)
+    }
+}

Conjuga/ConjugaUITests/VocabGridTests.swift

@@ -0,0 +1,53 @@
+import XCTest
+
+final class VocabGridTests: XCTestCase {
+
+    override func setUpWithError() throws {
+        continueAfterFailure = false
+    }
+
+    /// Verifies the chapter reader renders vocab tables as a paired Spanish↔English grid.
+    func testChapter4VocabGrid() throws {
+        let app = XCUIApplication()
+        app.launchArguments += ["-onboardingComplete", "YES"]
+        app.launch()
+
+        app.tabBars.buttons["Course"].tap()
+
+        let textbookRow = app.buttons.containing(NSPredicate(
+            format: "label CONTAINS[c] 'Complete Spanish'"
+        )).firstMatch
+        XCTAssertTrue(textbookRow.waitForExistence(timeout: 5))
+        textbookRow.tap()
+
+        let ch4 = app.buttons["textbook-chapter-row-4"]
+        XCTAssertTrue(ch4.waitForExistence(timeout: 3))
+        ch4.tap()
+
+        attach(app, name: "01-ch4-top")
+
+        // Tap the first vocab disclosure — "Vocabulary (N items)"
+        let vocabButton = app.buttons.matching(NSPredicate(
+            format: "label BEGINSWITH 'Vocabulary ('"
+        )).firstMatch
+        XCTAssertTrue(vocabButton.waitForExistence(timeout: 3))
+        vocabButton.tap()
+        Thread.sleep(forTimeInterval: 0.4)
+
+        attach(app, name: "02-ch4-vocab-open")
+
+        // Scroll a little and screenshot a deeper vocab — numbers table is
+        // typically a few screens down in chapter 4.
+        app.swipeUp(velocity: .fast)
+        app.swipeUp(velocity: .fast)
+        attach(app, name: "03-ch4-deeper")
+    }
+
+    private func attach(_ app: XCUIApplication, name: String) {
+        let s = app.screenshot()
+        let a = XCTAttachment(screenshot: s)
+        a.name = name
+        a.lifetime = .keepAlways
+        add(a)
+    }
+}

Conjuga/ConjugaWidget/CombinedWidget.swift

@@ -41,14 +41,16 @@ struct CombinedProvider: TimelineProvider {
     private func fetchWordOfDay(for date: Date) -> WordOfDay? {
         guard let localURL = SharedStore.localStoreURL() else { return nil }
 
-        // MUST declare all 6 local entities to match the main app's schema.
+        // MUST declare all 7 local entities to match the main app's schema.
-        // Declaring a subset would cause SwiftData to destructively migrate the store
+        // Declaring a subset would cause SwiftData to destructively migrate the
-        // on open, dropping the entities not listed here.
+        // store on open, dropping the entities not listed here (this is how we
+        // previously lost all TextbookChapter rows on every widget refresh).
         let config = ModelConfiguration(
             "local",
             schema: Schema([
                 Verb.self, VerbForm.self, IrregularSpan.self,
                 TenseGuide.self, CourseDeck.self, VocabCard.self,
+                TextbookChapter.self,
             ]),
             url: localURL,
             cloudKitDatabase: .none
@@ -56,6 +58,7 @@ struct CombinedProvider: TimelineProvider {
         guard let container = try? ModelContainer(
             for: Verb.self, VerbForm.self, IrregularSpan.self,
                  TenseGuide.self, CourseDeck.self, VocabCard.self,
+                 TextbookChapter.self,
             configurations: config
         ) else { return nil }
 

Conjuga/ConjugaWidget/WordOfDayWidget.swift

@@ -32,14 +32,16 @@ struct WordOfDayProvider: TimelineProvider {
     private func fetchWordOfDay(for date: Date) -> WordOfDay? {
         guard let localURL = SharedStore.localStoreURL() else { return nil }
 
-        // MUST declare all 6 local entities to match the main app's schema.
+        // MUST declare all 7 local entities to match the main app's schema.
-        // Declaring a subset would cause SwiftData to destructively migrate the store
+        // Declaring a subset would cause SwiftData to destructively migrate the
-        // on open, dropping the entities not listed here.
+        // store on open, dropping the entities not listed here (this is how we
+        // previously lost all TextbookChapter rows on every widget refresh).
         let config = ModelConfiguration(
             "local",
             schema: Schema([
                 Verb.self, VerbForm.self, IrregularSpan.self,
                 TenseGuide.self, CourseDeck.self, VocabCard.self,
+                TextbookChapter.self,
             ]),
             url: localURL,
             cloudKitDatabase: .none
@@ -47,6 +49,7 @@ struct WordOfDayProvider: TimelineProvider {
         guard let container = try? ModelContainer(
             for: Verb.self, VerbForm.self, IrregularSpan.self,
                  TenseGuide.self, CourseDeck.self, VocabCard.self,
+                 TextbookChapter.self,
             configurations: config
         ) else { return nil }
 

Conjuga/Scripts/textbook/build_book.py

@@ -0,0 +1,374 @@
+#!/usr/bin/env python3
+"""Merge chapters.json + answers.json + ocr.json → book.json (single source).
+
+Also emits vocab_cards.json: flashcards derived from vocab_image blocks where
+OCR text parses as a clean two-column (Spanish ↔ English) table.
+"""
+
+import json
+import re
+import sys
+from pathlib import Path
+
+HERE = Path(__file__).resolve().parent
+CHAPTERS_JSON = HERE / "chapters.json"
+ANSWERS_JSON = HERE / "answers.json"
+OCR_JSON = HERE / "ocr.json"
+OUT_BOOK = HERE / "book.json"
+OUT_VOCAB = HERE / "vocab_cards.json"
+
+COURSE_NAME = "Complete Spanish Step-by-Step"
+
+# Heuristic: parseable "Spanish | English" vocab rows.
+# OCR usually produces "word  —  translation" or "word translation" separated
+# by 2+ spaces. We detect rows that contain both Spanish and English words.
+SPANISH_ACCENT_RE = re.compile(r"[áéíóúñüÁÉÍÓÚÑÜ¿¡]")
+SPANISH_ARTICLES = {"el", "la", "los", "las", "un", "una", "unos", "unas"}
+ENGLISH_STARTERS = {"the", "a", "an", "to", "my", "his", "her", "our", "their", "your", "some"}
+# English-only words that would never appear as Spanish
+ENGLISH_ONLY_WORDS = {"the", "he", "she", "it", "we", "they", "I", "is", "are", "was", "were",
+                     "been", "have", "has", "had", "will", "would", "should", "could"}
+SEP_RE = re.compile(r"[ \t]{2,}|\s[—–−-]\s")
+
+
+def classify_line(line: str) -> str:
+    """Return 'es', 'en', or 'unknown' for the dominant language of a vocab line."""
+    line = line.strip()
+    if not line:
+        return "unknown"
+    # Accent = definitely Spanish
+    if SPANISH_ACCENT_RE.search(line):
+        return "es"
+    first = line.split()[0].lower().strip(",.;:")
+    if first in SPANISH_ARTICLES:
+        return "es"
+    if first in ENGLISH_STARTERS:
+        return "en"
+    # Check if the leading word is an English-only function word
+    if first in ENGLISH_ONLY_WORDS:
+        return "en"
+    return "unknown"
+
+
+def looks_english(word: str) -> bool:
+    """Legacy helper — kept for try_split_row below."""
+    w = word.lower().strip()
+    if not w:
+        return False
+    if SPANISH_ACCENT_RE.search(w):
+        return False
+    if w in SPANISH_ARTICLES:
+        return False
+    if w in ENGLISH_STARTERS or w in ENGLISH_ONLY_WORDS:
+        return True
+    return bool(re.match(r"^[a-z][a-z\s'/()\-,.]*$", w))
+
+
+def try_split_row(line: str) -> "tuple[str, str] | None":
+    """Split a line into (spanish, english) if it looks like a vocab entry."""
+    line = line.strip()
+    if not line or len(line) < 3:
+        return None
+    # Try explicit separators first
+    parts = SEP_RE.split(line)
+    parts = [p.strip() for p in parts if p.strip()]
+    if len(parts) == 2:
+        spanish, english = parts
+        if looks_english(english) and not looks_english(spanish.split()[0]):
+            return (spanish, english)
+    return None
+
+
+def load(p: Path) -> dict:
+    return json.loads(p.read_text(encoding="utf-8"))
+
+
+def build_vocab_cards_for_block(block: dict, ocr_entry: dict, chapter: dict, context_title: str, idx: int) -> list:
+    """Given a vocab_image block + its OCR lines, derive flashcards.
+
+    Vision OCR reads top-to-bottom, left-to-right; a two-column vocab table
+    produces Spanish lines first, then English lines. We split the list in
+    half when one side is predominantly Spanish and the other English.
+    Per-line '—' separators are also supported as a fallback.
+    """
+    cards = []
+    if not ocr_entry:
+        return cards
+    lines = [l.strip() for l in ocr_entry.get("lines", []) if l.strip()]
+    if not lines:
+        return cards
+
+    def card(front: str, back: str) -> dict:
+        return {
+            "front": front,
+            "back": back,
+            "chapter": chapter["number"],
+            "chapterTitle": chapter["title"],
+            "section": context_title,
+            "sourceImage": block["src"],
+        }
+
+    # Attempt 1: explicit inline separator (e.g. "la casa — the house")
+    inline = []
+    all_inline = True
+    for line in lines:
+        pair = try_split_row(line)
+        if pair:
+            inline.append(pair)
+        else:
+            all_inline = False
+            break
+    if all_inline and inline:
+        for es, en in inline:
+            cards.append(card(es, en))
+        return cards
+
+    # Attempt 2: block-alternating layout.
+    # Vision OCR reads columns top-to-bottom, so a 2-col table rendered across
+    # 2 visual columns produces runs like: [ES...ES][EN...EN][ES...ES][EN...EN]
+    # We classify each line, smooth "unknown" using neighbors, then pair
+    # same-sized consecutive ES/EN blocks.
+    classes = [classify_line(l) for l in lines]
+
+    # Pass 1: fill unknowns using nearest non-unknown neighbor (forward)
+    last_known = "unknown"
+    forward = []
+    for c in classes:
+        if c != "unknown":
+            last_known = c
+        forward.append(last_known)
+    # Pass 2: backfill leading unknowns (backward)
+    last_known = "unknown"
+    backward = [""] * len(classes)
+    for i in range(len(classes) - 1, -1, -1):
+        if classes[i] != "unknown":
+            last_known = classes[i]
+        backward[i] = last_known
+    # Merge: prefer forward unless still unknown
+    resolved = []
+    for f, b in zip(forward, backward):
+        if f != "unknown":
+            resolved.append(f)
+        elif b != "unknown":
+            resolved.append(b)
+        else:
+            resolved.append("unknown")
+
+    # Group consecutive same-lang lines
+    blocks: list = []
+    cur_lang: "str | None" = None
+    cur_block: list = []
+    for line, lang in zip(lines, resolved):
+        if lang != cur_lang:
+            if cur_block and cur_lang is not None:
+                blocks.append((cur_lang, cur_block))
+            cur_block = [line]
+            cur_lang = lang
+        else:
+            cur_block.append(line)
+    if cur_block and cur_lang is not None:
+        blocks.append((cur_lang, cur_block))
+
+    # Walk blocks pairing ES then EN of equal length
+    i = 0
+    while i < len(blocks) - 1:
+        lang_a, lines_a = blocks[i]
+        lang_b, lines_b = blocks[i + 1]
+        if lang_a == "es" and lang_b == "en" and len(lines_a) == len(lines_b):
+            for es, en in zip(lines_a, lines_b):
+                cards.append(card(es, en))
+            i += 2
+            continue
+        # If reversed order (some pages have EN column on left), try that too
+        if lang_a == "en" and lang_b == "es" and len(lines_a) == len(lines_b):
+            for es, en in zip(lines_b, lines_a):
+                cards.append(card(es, en))
+            i += 2
+            continue
+        i += 1
+
+    return cards
+
+
+def clean_instruction(text: str) -> str:
+    """Strip leading/trailing emphasis markers from a parsed instruction."""
+    # Our XHTML parser emitted * and ** for emphasis; flatten them
+    t = re.sub(r"\*+", "", text)
+    return t.strip()
+
+
+def merge() -> None:
+    chapters_data = load(CHAPTERS_JSON)
+    answers_data = load(ANSWERS_JSON)
+    try:
+        ocr_data = load(OCR_JSON)
+    except FileNotFoundError:
+        print("ocr.json not found — proceeding with empty OCR data")
+        ocr_data = {}
+
+    answers = answers_data["answers"]
+    chapters = chapters_data["chapters"]
+    parts = chapters_data.get("part_memberships", {})
+
+    book_chapters = []
+    all_vocab_cards = []
+    missing_ocr = set()
+    current_section_title = ""
+
+    for ch in chapters:
+        out_blocks = []
+        current_section_title = ch["title"]
+
+        for bi, block in enumerate(ch["blocks"]):
+            k = block["kind"]
+
+            if k == "heading":
+                current_section_title = block["text"]
+                out_blocks.append(block)
+                continue
+
+            if k == "paragraph":
+                out_blocks.append(block)
+                continue
+
+            if k == "key_vocab_header":
+                out_blocks.append(block)
+                continue
+
+            if k == "vocab_image":
+                ocr_entry = ocr_data.get(block["src"])
+                if ocr_entry is None:
+                    missing_ocr.add(block["src"])
+                derived = build_vocab_cards_for_block(
+                    block, ocr_entry, ch, current_section_title, bi
+                )
+                all_vocab_cards.extend(derived)
+                out_blocks.append({
+                    "kind": "vocab_table",
+                    "sourceImage": block["src"],
+                    "ocrLines": ocr_entry.get("lines", []) if ocr_entry else [],
+                    "ocrConfidence": ocr_entry.get("confidence", 0.0) if ocr_entry else 0.0,
+                    "cardCount": len(derived),
+                })
+                continue
+
+            if k == "exercise":
+                ans = answers.get(block["id"])
+                image_ocr_lines = []
+                for src in block.get("image_refs", []):
+                    e = ocr_data.get(src)
+                    if e is None:
+                        missing_ocr.add(src)
+                        continue
+                    image_ocr_lines.extend(e.get("lines", []))
+
+                # Build the final prompt list. If we have text prompts from
+                # XHTML, prefer them. Otherwise, attempt to use OCR lines.
+                prompts = [p for p in block.get("prompts", []) if p.strip()]
+                extras = [e for e in block.get("extra", []) if e.strip()]
+                if not prompts and image_ocr_lines:
+                    # Extract numbered lines from OCR (look for "1.  ..." pattern)
+                    for line in image_ocr_lines:
+                        m = re.match(r"^(\d+)[.)]\s*(.+)", line.strip())
+                        if m:
+                            prompts.append(f"{m.group(1)}. {m.group(2)}")
+
+                # Cross-reference prompts with answers
+                sub = ans["subparts"] if ans else []
+                answer_items = []
+                for sp in sub:
+                    for it in sp["items"]:
+                        answer_items.append({
+                            "label": sp["label"],
+                            "number": it["number"],
+                            "answer": it["answer"],
+                            "alternates": it["alternates"],
+                        })
+
+                out_blocks.append({
+                    "kind": "exercise",
+                    "id": block["id"],
+                    "ansAnchor": block.get("ans_anchor", ""),
+                    "instruction": clean_instruction(block.get("instruction", "")),
+                    "extra": extras,
+                    "prompts": prompts,
+                    "ocrLines": image_ocr_lines,
+                    "freeform": ans["freeform"] if ans else False,
+                    "answerItems": answer_items,
+                    "answerRaw": ans["raw"] if ans else "",
+                    "answerSubparts": sub,
+                })
+                continue
+
+            out_blocks.append(block)
+
+        book_chapters.append({
+            "id": ch["id"],
+            "number": ch["number"],
+            "title": ch["title"],
+            "part": ch.get("part"),
+            "blocks": out_blocks,
+        })
+
+    book = {
+        "courseName": COURSE_NAME,
+        "totalChapters": len(book_chapters),
+        "totalExercises": sum(
+            1 for ch in book_chapters for b in ch["blocks"] if b["kind"] == "exercise"
+        ),
+        "totalVocabTables": sum(
+            1 for ch in book_chapters for b in ch["blocks"] if b["kind"] == "vocab_table"
+        ),
+        "totalVocabCards": len(all_vocab_cards),
+        "parts": parts,
+        "chapters": book_chapters,
+    }
+    OUT_BOOK.write_text(json.dumps(book, ensure_ascii=False))
+
+    # Vocab cards as a separate file (grouped per chapter so they can be seeded
+    # as CourseDecks in the existing schema).
+    vocab_by_chapter: dict = {}
+    for card in all_vocab_cards:
+        vocab_by_chapter.setdefault(card["chapter"], []).append(card)
+    OUT_VOCAB.write_text(json.dumps({
+        "courseName": COURSE_NAME,
+        "chapters": [
+            {
+                "chapter": ch_num,
+                "cards": cards,
+            }
+            for ch_num, cards in sorted(vocab_by_chapter.items())
+        ],
+    }, ensure_ascii=False, indent=2))
+
+    # Summary
+    print(f"Wrote {OUT_BOOK}")
+    print(f"Wrote {OUT_VOCAB}")
+    print(f"Chapters:           {book['totalChapters']}")
+    print(f"Exercises:          {book['totalExercises']}")
+    print(f"Vocab tables:       {book['totalVocabTables']}")
+    print(f"Vocab cards (auto): {book['totalVocabCards']}")
+    if missing_ocr:
+        print(f"Missing OCR for {len(missing_ocr)} images (first 5): {sorted(list(missing_ocr))[:5]}")
+
+    # Validation
+    total_exercises = book["totalExercises"]
+    exercises_with_prompts = sum(
+        1 for ch in book_chapters for b in ch["blocks"]
+        if b["kind"] == "exercise" and (b["prompts"] or b["extra"])
+    )
+    exercises_with_answers = sum(
+        1 for ch in book_chapters for b in ch["blocks"]
+        if b["kind"] == "exercise" and b["answerItems"]
+    )
+    exercises_freeform = sum(
+        1 for ch in book_chapters for b in ch["blocks"]
+        if b["kind"] == "exercise" and b["freeform"]
+    )
+    print(f"Exercises with prompts: {exercises_with_prompts}/{total_exercises}")
+    print(f"Exercises with answers: {exercises_with_answers}/{total_exercises}")
+    print(f"Freeform exercises:     {exercises_freeform}")
+
+
+if __name__ == "__main__":
+    merge()

Conjuga/Scripts/textbook/build_review.py

@@ -0,0 +1,126 @@
+#!/usr/bin/env python3
+"""Render book.json + ocr.json into a static HTML review page.
+
+The HTML surfaces low-confidence OCR results in red, and shows the parsed
+exercise prompts/answers next to the original image. Designed for rapid
+visual diffing against the source book.
+"""
+
+import html
+import json
+from pathlib import Path
+
+HERE = Path(__file__).resolve().parent
+BOOK = HERE / "book.json"
+OCR = HERE / "ocr.json"
+OUT_HTML = HERE / "review.html"
+EPUB_IMAGES = Path(HERE).parents[2] / "epub_extract" / "OEBPS"
+IMAGE_REL = EPUB_IMAGES.relative_to(HERE.parent) if False else EPUB_IMAGES
+
+
+def load(p: Path) -> dict:
+    return json.loads(p.read_text(encoding="utf-8"))
+
+
+def esc(s: str) -> str:
+    return html.escape(s or "")
+
+
+def img_tag(src: str) -> str:
+    full = (EPUB_IMAGES / src).resolve()
+    return f'<img src="file://{full}" alt="{esc(src)}" class="src"/>'
+
+
+def render() -> None:
+    book = load(BOOK)
+    ocr = load(OCR) if OCR.exists() else {}
+
+    out: list = []
+    out.append("""<!DOCTYPE html>
+<html><head><meta charset='utf-8'><title>Book review</title>
+<style>
+body { font-family: -apple-system, system-ui, sans-serif; margin: 2em; max-width: 1000px; color: #222; }
+h1 { color: #c44; }
+h2.chapter { background: #eee; padding: 0.5em; border-left: 4px solid #c44; }
+h3.heading { color: #555; }
+.para { margin: 0.5em 0; }
+.vocab-table { background: #fafff0; padding: 0.5em; margin: 0.5em 0; border: 1px solid #bda; border-radius: 6px; }
+.ocr-line { font-family: ui-monospace, monospace; font-size: 12px; }
+.lowconf { color: #c44; background: #fee; }
+.exercise { background: #fff8e8; padding: 0.5em; margin: 0.75em 0; border: 1px solid #cb9; border-radius: 6px; }
+.prompt { font-family: ui-monospace, monospace; font-size: 13px; margin: 2px 0; }
+.answer { color: #080; font-family: ui-monospace, monospace; font-size: 13px; }
+img.src { max-width: 520px; border: 1px solid #ccc; margin: 4px 0; }
+.kv { color: #04a; font-weight: bold; }
+summary { cursor: pointer; font-weight: bold; color: #666; }
+.card-pair { font-family: ui-monospace, monospace; font-size: 12px; }
+.card-es { color: #04a; }
+.card-en { color: #555; }
+.counts { color: #888; font-size: 12px; }
+</style></head><body>""")
+    out.append(f"<h1>{esc(book['courseName'])} — review</h1>")
+    out.append(f"<p>{book['totalChapters']} chapters · {book['totalExercises']} exercises · {book['totalVocabTables']} vocab tables · {book['totalVocabCards']} auto-derived cards</p>")
+
+    for ch in book["chapters"]:
+        part = ch.get("part")
+        part_str = f" (Part {part})" if part else ""
+        out.append(f"<h2 class='chapter'>Chapter {ch['number']}: {esc(ch['title'])}{esc(part_str)}</h2>")
+
+        for b in ch["blocks"]:
+            kind = b["kind"]
+            if kind == "heading":
+                level = b["level"]
+                out.append(f"<h{level} class='heading'>{esc(b['text'])}</h{level}>")
+            elif kind == "paragraph":
+                out.append(f"<p class='para'>{esc(b['text'])}</p>")
+            elif kind == "key_vocab_header":
+                out.append(f"<p class='kv'>★ Key Vocabulary</p>")
+            elif kind == "vocab_table":
+                src = b["sourceImage"]
+                conf = b["ocrConfidence"]
+                conf_class = "lowconf" if conf < 0.85 else ""
+                out.append(f"<div class='vocab-table'>")
+                out.append(f"<details><summary>vocab {esc(src)} · confidence {conf:.2f} · {b['cardCount']} card(s)</summary>")
+                out.append(img_tag(src))
+                out.append("<div>")
+                for line in b.get("ocrLines", []):
+                    out.append(f"<div class='ocr-line {conf_class}'>{esc(line)}</div>")
+                out.append("</div>")
+                # Show derived pairs (if any). We don't have them inline in book.json,
+                # but we can recompute from ocrLines using the same function.
+                out.append("</details></div>")
+            elif kind == "exercise":
+                out.append(f"<div class='exercise'>")
+                out.append(f"<b>Exercise {esc(b['id'])}</b> — <i>{esc(b['instruction'])}</i>")
+                if b.get("extra"):
+                    for e in b["extra"]:
+                        out.append(f"<div class='para'>{esc(e)}</div>")
+                if b.get("ocrLines"):
+                    out.append(f"<details><summary>OCR lines from image</summary>")
+                    for line in b["ocrLines"]:
+                        out.append(f"<div class='ocr-line'>{esc(line)}</div>")
+                    out.append("</details>")
+                if b.get("prompts"):
+                    out.append("<div><b>Parsed prompts:</b></div>")
+                    for p in b["prompts"]:
+                        out.append(f"<div class='prompt'>• {esc(p)}</div>")
+                if b.get("answerItems"):
+                    out.append("<div><b>Answer key:</b></div>")
+                    for a in b["answerItems"]:
+                        label_str = f"{a['label']}. " if a.get("label") else ""
+                        alts = ", ".join(a["alternates"])
+                        alt_str = f"  <span style='color:#999'>(also: {esc(alts)})</span>" if alts else ""
+                        out.append(f"<div class='answer'>{esc(label_str)}{a['number']}. {esc(a['answer'])}{alt_str}</div>")
+                if b.get("freeform"):
+                    out.append("<div style='color:#c44'>(Freeform — answers will vary)</div>")
+                for img_src in b.get("image_refs", []):
+                    out.append(img_tag(img_src))
+                out.append("</div>")
+
+    out.append("</body></html>")
+    OUT_HTML.write_text("\n".join(out), encoding="utf-8")
+    print(f"Wrote {OUT_HTML}")
+
+
+if __name__ == "__main__":
+    render()

Conjuga/Scripts/textbook/extract_answers.py

@@ -0,0 +1,205 @@
+#!/usr/bin/env python3
+"""Parse ans.xhtml into structured answers.json.
+
+Output schema:
+{
+  "answers": {
+    "1.1": {
+      "id": "1.1",
+      "anchor": "ch1ans1",
+      "chapter": 1,
+      "subparts": [
+        {"label": null, "items": [
+          {"number": 1, "answer": "el", "alternates": []},
+          {"number": 2, "answer": "el", "alternates": []},
+          ...
+        ]}
+      ],
+      "freeform": false,        # true if "Answers will vary"
+      "raw": "..."              # raw text for fallback
+    },
+    "2.4": {                     # multi-part exercise
+      "subparts": [
+        {"label": "A", "items": [...]},
+        {"label": "B", "items": [...]},
+        {"label": "C", "items": [...]}
+      ]
+    }
+  }
+}
+"""
+
+import json
+import re
+from pathlib import Path
+from bs4 import BeautifulSoup, NavigableString
+
+ROOT = Path(__file__).resolve().parents[3] / "epub_extract" / "OEBPS"
+OUT = Path(__file__).resolve().parent / "answers.json"
+
+ANSWER_CLASSES = {"answerq", "answerq1", "answerq2", "answerqa"}
+EXERCISE_ID_RE = re.compile(r"^([0-9]+)\.([0-9]+)$")
+SUBPART_LABEL_RE = re.compile(r"^([A-Z])\b")
+NUMBERED_ITEM_RE = re.compile(r"(?:^|\s)(\d+)\.\s+")
+FREEFORM_PATTERNS = [
+    re.compile(r"answers? will vary", re.IGNORECASE),
+    re.compile(r"answer will vary", re.IGNORECASE),
+]
+OR_TOKEN = "{{OR}}"
+
+
+def render_with_or(p) -> str:
+    """Convert <p> to plain text, replacing 'OR' span markers with sentinel."""
+    soup = BeautifulSoup(str(p), "lxml")
+    # Replace <span class="small">OR</span> with sentinel
+    for span in soup.find_all("span"):
+        cls = span.get("class") or []
+        if "small" in cls and span.get_text(strip=True).upper() == "OR":
+            span.replace_with(f" {OR_TOKEN} ")
+    # Drop pagebreak spans
+    for span in soup.find_all("span", attrs={"epub:type": "pagebreak"}):
+        span.decompose()
+    # Drop emphasis but keep text
+    for tag in soup.find_all(["em", "i", "strong", "b"]):
+        tag.unwrap()
+    text = soup.get_text(separator=" ", strip=False)
+    text = re.sub(r"\s+", " ", text).strip()
+    return text
+
+
+def split_numbered_items(text: str) -> "list[dict]":
+    """Given '1. el 2. la 3. el ...' return [{'number':1,'answer':'el'}, ...]."""
+    # Find positions of N. tokens
+    matches = list(NUMBERED_ITEM_RE.finditer(text))
+    items = []
+    for i, m in enumerate(matches):
+        num = int(m.group(1))
+        start = m.end()
+        end = matches[i + 1].start() if i + 1 < len(matches) else len(text)
+        body = text[start:end].strip().rstrip(".,;")
+        # Split alternates on the OR token
+        parts = [p.strip() for p in body.split(OR_TOKEN) if p.strip()]
+        if not parts:
+            continue
+        items.append({
+            "number": num,
+            "answer": parts[0],
+            "alternates": parts[1:],
+        })
+    return items
+
+
+def parse_subpart_label(text: str) -> "tuple[str | None, str]":
+    """Try to peel a leading subpart label (A, B, C) from the text.
+    Returns (label_or_None, remaining_text)."""
+    # Pattern at start: "A " or "A      " (lots of whitespace from <em>A</em><tab>)
+    m = re.match(r"^([A-Z])\s+(?=\d)", text)
+    if m:
+        return m.group(1), text[m.end():]
+    return None, text
+
+
+def parse_answer_paragraph(p, exercise_id: str) -> "list[dict]":
+    """Convert one <p> into a list of subparts.
+    For p.answerq, the text typically starts with the exercise id, then items.
+    For p.answerqa, the text starts with a subpart label letter."""
+    raw = render_with_or(p)
+    # Strip the leading exercise id if present
+    raw = re.sub(rf"^{re.escape(exercise_id)}\s*", "", raw)
+
+    label, body = parse_subpart_label(raw)
+
+    # Detect freeform
+    freeform = any(pat.search(body) for pat in FREEFORM_PATTERNS)
+    if freeform:
+        return [{"label": label, "items": [], "freeform": True, "raw": body}]
+
+    items = split_numbered_items(body)
+    return [{"label": label, "items": items, "freeform": False, "raw": body}]
+
+
+def main() -> None:
+    src = ROOT / "ans.xhtml"
+    soup = BeautifulSoup(src.read_text(encoding="utf-8"), "lxml")
+    body = soup.find("body")
+
+    answers: dict = {}
+    current_chapter = None
+    current_exercise_id: "str | None" = None
+
+    for el in body.find_all(["h3", "p"]):
+        classes = set(el.get("class") or [])
+
+        # Chapter boundary
+        if el.name == "h3" and "h3b" in classes:
+            text = el.get_text(strip=True)
+            m = re.search(r"Chapter\s+(\d+)", text)
+            if m:
+                current_chapter = int(m.group(1))
+                current_exercise_id = None
+            continue
+
+        if el.name != "p" or not (classes & ANSWER_CLASSES):
+            continue
+
+        # Find the exercise-id anchor (only present on p.answerq, not on continuation)
+        a = el.find("a", href=True)
+        ex_link = None
+        if a:
+            link_text = a.get_text(strip=True)
+            if EXERCISE_ID_RE.match(link_text):
+                ex_link = link_text
+
+        if ex_link:
+            current_exercise_id = ex_link
+            anchor = ""
+            href = a.get("href", "")
+            anchor_m = re.search(r"#(ch\d+ans\d+)", href + " " + (a.get("id") or ""))
+            anchor = anchor_m.group(1) if anchor_m else (a.get("id") or "")
+            # Use the anchor's `id` attr if it's the entry id (e.g. "ch1ans1")
+            entry_id = a.get("id") or anchor
+
+            answers[ex_link] = {
+                "id": ex_link,
+                "anchor": entry_id,
+                "chapter": current_chapter,
+                "subparts": [],
+                "freeform": False,
+                "raw": "",
+            }
+            new_subparts = parse_answer_paragraph(el, ex_link)
+            answers[ex_link]["subparts"].extend(new_subparts)
+            answers[ex_link]["raw"] = render_with_or(el)
+            answers[ex_link]["freeform"] = any(sp["freeform"] for sp in new_subparts)
+        else:
+            # Continuation paragraph for current exercise
+            if current_exercise_id and current_exercise_id in answers:
+                more = parse_answer_paragraph(el, current_exercise_id)
+                answers[current_exercise_id]["subparts"].extend(more)
+                if any(sp["freeform"] for sp in more):
+                    answers[current_exercise_id]["freeform"] = True
+
+    out = {"answers": answers}
+    OUT.write_text(json.dumps(out, ensure_ascii=False, indent=2))
+
+    total = len(answers)
+    freeform = sum(1 for v in answers.values() if v["freeform"])
+    multipart = sum(1 for v in answers.values() if len(v["subparts"]) > 1)
+    total_items = sum(
+        len(sp["items"]) for v in answers.values() for sp in v["subparts"]
+    )
+    with_alternates = sum(
+        1 for v in answers.values()
+        for sp in v["subparts"] for it in sp["items"]
+        if it["alternates"]
+    )
+    print(f"Exercises with answers: {total}")
+    print(f"  freeform:             {freeform}")
+    print(f"  multi-part (A/B/C):   {multipart}")
+    print(f"  total numbered items: {total_items}")
+    print(f"  items with alternates:{with_alternates}")
+    print(f"Wrote {OUT}")
+
+
+if __name__ == "__main__":
+    main()

Conjuga/Scripts/textbook/extract_chapters.py

@@ -0,0 +1,369 @@
+#!/usr/bin/env python3
+"""Parse all chapter XHTMLs + appendix into structured chapters.json.
+
+Output schema:
+{
+  "chapters": [
+    {
+      "id": "ch1",
+      "number": 1,
+      "title": "Nouns, Articles, and Adjectives",
+      "part": 1,                          # part 1/2/3 or null
+      "blocks": [                         # ordered content
+        {"kind": "heading", "level": 3, "text": "..."},
+        {"kind": "paragraph", "text": "...", "hasItalic": false},
+        {"kind": "key_vocab_header", "title": "Los colores (The colors)"},
+        {"kind": "vocab_image", "src": "f0010-03.jpg"},
+        {
+          "kind": "exercise",
+          "id": "1.1",
+          "ans_anchor": "ch1ans1",
+          "instruction": "Write the appropriate...",
+          "image_refs": ["f0005-02.jpg"]
+        },
+        {"kind": "image", "src": "...", "alt": "..."}
+      ]
+    }
+  ]
+}
+"""
+
+import json
+import re
+from pathlib import Path
+from bs4 import BeautifulSoup
+
+ROOT = Path(__file__).resolve().parents[3] / "epub_extract" / "OEBPS"
+OUT = Path(__file__).resolve().parent / "chapters.json"
+
+# Common icon images embedded in headings — ignore when collecting content images
+ICON_IMAGES = {"Common01.jpg", "Common02.jpg", "Common03.jpg", "Common04.jpg", "Common05.jpg"}
+
+EXERCISE_ID_RE = re.compile(r"Exercise\s+([0-9]+\.[0-9]+)")
+ANS_REF_RE = re.compile(r"ch(\d+)ans(\d+)")
+
+
+def clean_text(el) -> str:
+    """Extract text preserving inline emphasis markers."""
+    if el is None:
+        return ""
+    # Replace <em>/<i> with markdown-ish *...*, <strong>/<b> with **...**
+    html = str(el)
+    soup = BeautifulSoup(html, "lxml")
+    # First: flatten nested emphasis so we don't emit overlapping markers.
+    # For <strong><em>X</em></strong>, drop the inner em (the bold wrapping
+    # already carries the emphasis visually). Same for <em><strong>...</strong></em>.
+    for tag in soup.find_all(["strong", "b"]):
+        for inner in tag.find_all(["em", "i"]):
+            inner.unwrap()
+    for tag in soup.find_all(["em", "i"]):
+        for inner in tag.find_all(["strong", "b"]):
+            inner.unwrap()
+    # Drop ALL inline emphasis. The source has nested/sibling em/strong
+    # patterns that CommonMark can't reliably parse, causing markers to leak
+    # into the UI. Plain text renders cleanly everywhere.
+    for tag in soup.find_all(["em", "i", "strong", "b"]):
+        tag.unwrap()
+    # Drop pagebreak spans
+    for tag in soup.find_all("span", attrs={"epub:type": "pagebreak"}):
+        tag.decompose()
+    # Replace <br/> with newline
+    for br in soup.find_all("br"):
+        br.replace_with("\n")
+    # Use a separator so adjacent inline tags don't concatenate without spaces
+    # (e.g. "<strong><em>Ir</em></strong> and" would otherwise become "Irand").
+    text = soup.get_text(separator=" ", strip=False)
+    # Collapse runs of whitespace first.
+    text = re.sub(r"\s+", " ", text).strip()
+    # Strip any stray asterisks that sneak through (e.g. author's literal *).
+    text = text.replace("*", "")
+    # De-space punctuation
+    text = re.sub(r"\s+([,.;:!?])", r"\1", text)
+    # Tighten brackets that picked up separator-spaces: "( foo )" -> "(foo)"
+    text = re.sub(r"([(\[])\s+", r"\1", text)
+    text = re.sub(r"\s+([)\]])", r"\1", text)
+    # Collapse any double-spaces
+    text = re.sub(r"  +", " ", text).strip()
+    return text
+
+
+def is_exercise_header(h) -> bool:
+    """Heading with an <a href='ans.xhtml#...'>Exercise N.N</a> link.
+    Chapters 1-16 use h3.h3k; chapters 17+ use h4.h4."""
+    if h.name not in ("h3", "h4"):
+        return False
+    a = h.find("a", href=True)
+    if a and "ans.xhtml" in a["href"]:
+        return True
+    return False
+
+
+def is_key_vocab_header(h) -> bool:
+    """Heading with 'Key Vocabulary' text (no anchor link to answers)."""
+    if h.name not in ("h3", "h4"):
+        return False
+    text = h.get_text(strip=True)
+    if "Key Vocabulary" in text and not h.find("a", href=lambda v: v and "ans.xhtml" in v):
+        return True
+    return False
+
+
+def extract_image_srcs(parent) -> list:
+    """Return list of image src attributes, skipping icon images."""
+    srcs = []
+    for img in parent.find_all("img"):
+        src = img.get("src", "")
+        if not src or Path(src).name in ICON_IMAGES:
+            continue
+        srcs.append(src)
+    return srcs
+
+
+def parse_chapter(path: Path) -> "dict | None":
+    """Parse one chapter file into structured blocks."""
+    html = path.read_text(encoding="utf-8")
+    soup = BeautifulSoup(html, "lxml")
+    body = soup.find("body")
+    if body is None:
+        return None
+
+    # Chapter number + title
+    number = None
+    title = ""
+    h2s = body.find_all("h2")
+    for h2 in h2s:
+        classes = h2.get("class") or []
+        # Use a separator so consecutive inline tags don't concatenate
+        # (e.g. "<strong><em>Ir</em></strong> and the Future" → "Ir and the Future")
+        text_with_sep = re.sub(r"\s+", " ", h2.get_text(" ", strip=True))
+        # Strip spaces that were inserted before punctuation
+        text_with_sep = re.sub(r"\s+([,.;:!?])", r"\1", text_with_sep).strip()
+        if "h2c" in classes and text_with_sep.isdigit():
+            number = int(text_with_sep)
+        # Chapters 1–16 use h2c1; chapters 17+ use h2-c
+        elif ("h2c1" in classes or "h2-c" in classes) and not title:
+            title = text_with_sep
+    if number is None:
+        # Try id on chapter header (ch1 → 1)
+        for h2 in h2s:
+            id_ = h2.get("id", "")
+            m = re.match(r"ch(\d+)", id_)
+            if m:
+                number = int(m.group(1))
+                break
+
+    chapter_id = path.stem  # ch1, ch2, ...
+
+    # Walk section content in document order
+    section = body.find("section") or body
+    blocks: list = []
+    pending_instruction = None  # holds italic paragraph following an exercise header
+
+    for el in section.descendants:
+        if el.name is None:
+            continue
+
+        classes = el.get("class") or []
+
+        # Skip nested tags already captured via parent processing
+        # We operate only on direct h2/h3/h4/h5/p elements
+        if el.name not in ("h2", "h3", "h4", "h5", "p"):
+            continue
+
+        # Exercise header detection (h3 in ch1-16, h4 in ch17+)
+        if is_exercise_header(el):
+            a = el.find("a", href=True)
+            href = a["href"] if a else ""
+            m = EXERCISE_ID_RE.search(el.get_text())
+            ex_id = m.group(1) if m else ""
+            anchor_m = ANS_REF_RE.search(href)
+            ans_anchor = anchor_m.group(0) if anchor_m else ""
+            blocks.append({
+                "kind": "exercise",
+                "id": ex_id,
+                "ans_anchor": ans_anchor,
+                "instruction": "",
+                "image_refs": [],
+                "prompts": []
+            })
+            pending_instruction = blocks[-1]
+            continue
+
+        # Key Vocabulary header
+        if is_key_vocab_header(el):
+            blocks.append({"kind": "key_vocab_header", "title": "Key Vocabulary"})
+            pending_instruction = None
+            continue
+
+        # Other headings
+        if el.name in ("h2", "h3", "h4", "h5"):
+            if el.name == "h2":
+                # Skip the chapter-number/chapter-title h2s we already captured
+                continue
+            txt = clean_text(el)
+            if txt:
+                blocks.append({
+                    "kind": "heading",
+                    "level": int(el.name[1]),
+                    "text": txt,
+                })
+            pending_instruction = None
+            continue
+
+        # Paragraphs
+        if el.name == "p":
+            imgs = extract_image_srcs(el)
+            text = clean_text(el)
+            p_classes = set(classes)
+
+            # Skip pure blank-line class ("nump" = underscore lines under number prompts)
+            if p_classes & {"nump", "numpa"} and not text:
+                continue
+
+            # Exercise prompt: <p class="number">1.  Prompt text</p>
+            # Also number1, number2 (continuation numbering), numbera, numbert
+            if pending_instruction is not None and p_classes & {"number", "number1", "number2", "numbera", "numbert"}:
+                if text:
+                    pending_instruction["prompts"].append(text)
+                continue
+
+            # Image container for a pending exercise
+            if pending_instruction is not None and imgs and not text:
+                pending_instruction["image_refs"].extend(imgs)
+                continue
+
+            # Instruction line right after the exercise header
+            if pending_instruction is not None and text and not imgs and not pending_instruction["instruction"]:
+                pending_instruction["instruction"] = text
+                continue
+
+            # While in pending-exercise state, extra text paragraphs are word
+            # banks / context ("from the following list:" etc) — keep pending alive.
+            if pending_instruction is not None and text and not imgs:
+                pending_instruction.setdefault("extra", []).append(text)
+                continue
+
+            # Paragraphs that contain an image belong to vocab/key-vocab callouts
+            if imgs and not text:
+                for src in imgs:
+                    blocks.append({"kind": "vocab_image", "src": src})
+                continue
+
+            # Mixed paragraph: image with caption
+            if imgs and text:
+                for src in imgs:
+                    blocks.append({"kind": "vocab_image", "src": src})
+                blocks.append({"kind": "paragraph", "text": text})
+                continue
+
+            # Plain paragraph — outside any exercise
+            if text:
+                blocks.append({"kind": "paragraph", "text": text})
+
+    return {
+        "id": chapter_id,
+        "number": number,
+        "title": title,
+        "blocks": blocks,
+    }
+
+
+def assign_parts(chapters: list, part_files: "dict[int, list[int]]") -> None:
+    """Annotate chapters with part number based on TOC membership."""
+    for part_num, chapter_nums in part_files.items():
+        for ch in chapters:
+            if ch["number"] in chapter_nums:
+                ch["part"] = part_num
+    for ch in chapters:
+        ch.setdefault("part", None)
+
+
+def read_part_memberships() -> "dict[int, list[int]]":
+    """Derive part→chapter grouping from the OPF spine order."""
+    opf = next(ROOT.glob("*.opf"), None)
+    if opf is None:
+        return {}
+    soup = BeautifulSoup(opf.read_text(encoding="utf-8"), "xml")
+    memberships: dict = {}
+    current_part: "int | None" = None
+    for item in soup.find_all("item"):
+        href = item.get("href", "")
+        m_part = re.match(r"part(\d+)\.xhtml", href)
+        m_ch = re.match(r"ch(\d+)\.xhtml", href)
+        if m_part:
+            current_part = int(m_part.group(1))
+            memberships.setdefault(current_part, [])
+        elif m_ch and current_part is not None:
+            memberships[current_part].append(int(m_ch.group(1)))
+    # Manifest order tends to match spine order for this book; verify via spine just in case
+    spine = soup.find("spine")
+    if spine is not None:
+        order = []
+        for ref in spine.find_all("itemref"):
+            idref = ref.get("idref")
+            item = soup.find("item", attrs={"id": idref})
+            if item is not None:
+                order.append(item.get("href", ""))
+        # Rebuild from spine order
+        memberships = {}
+        current_part = None
+        for href in order:
+            m_part = re.match(r"part(\d+)\.xhtml", href)
+            m_ch = re.match(r"ch(\d+)\.xhtml", href)
+            if m_part:
+                current_part = int(m_part.group(1))
+                memberships.setdefault(current_part, [])
+            elif m_ch and current_part is not None:
+                memberships[current_part].append(int(m_ch.group(1)))
+    return memberships
+
+
+def main() -> None:
+    chapter_files = sorted(
+        ROOT.glob("ch*.xhtml"),
+        key=lambda p: int(re.match(r"ch(\d+)", p.stem).group(1))
+    )
+    chapters = []
+    for path in chapter_files:
+        ch = parse_chapter(path)
+        if ch:
+            chapters.append(ch)
+
+    part_memberships = read_part_memberships()
+    assign_parts(chapters, part_memberships)
+
+    out = {
+        "chapters": chapters,
+        "part_memberships": part_memberships,
+    }
+    OUT.write_text(json.dumps(out, ensure_ascii=False, indent=2))
+
+    # Summary
+    ex_total = sum(1 for ch in chapters for b in ch["blocks"] if b["kind"] == "exercise")
+    ex_with_prompts = sum(
+        1 for ch in chapters for b in ch["blocks"]
+        if b["kind"] == "exercise" and b["prompts"]
+    )
+    ex_with_images = sum(
+        1 for ch in chapters for b in ch["blocks"]
+        if b["kind"] == "exercise" and b["image_refs"]
+    )
+    ex_empty = sum(
+        1 for ch in chapters for b in ch["blocks"]
+        if b["kind"] == "exercise" and not b["prompts"] and not b["image_refs"]
+    )
+    para_total = sum(1 for ch in chapters for b in ch["blocks"] if b["kind"] == "paragraph")
+    vocab_img_total = sum(1 for ch in chapters for b in ch["blocks"] if b["kind"] == "vocab_image")
+    print(f"Chapters:             {len(chapters)}")
+    print(f"Exercises total:      {ex_total}")
+    print(f"  with text prompts:  {ex_with_prompts}")
+    print(f"  with image prompts: {ex_with_images}")
+    print(f"  empty:              {ex_empty}")
+    print(f"Paragraphs:           {para_total}")
+    print(f"Vocab images:         {vocab_img_total}")
+    print(f"Parts: {part_memberships}")
+    print(f"Wrote {OUT}")
+
+
+if __name__ == "__main__":
+    main()

Conjuga/Scripts/textbook/extract_pdf_text.py

@@ -0,0 +1,94 @@
+#!/usr/bin/env python3
+"""Extract clean text from the PDF source and map each PDF page to the
+book's printed page number.
+
+Output: pdf_text.json
+{
+  "pdfPageCount": 806,
+  "bookPages": {
+    "3": { "text": "...", "pdfIndex": 29 },
+    "4": { ... },
+    ...
+  },
+  "unmapped": [list of pdfIndex values with no detectable book page number]
+}
+"""
+
+import json
+import re
+from pathlib import Path
+import pypdf
+
+HERE = Path(__file__).resolve().parent
+PDF = next(
+    Path(__file__).resolve().parents[3].glob("Complete Spanish Step-By-Step*.pdf"),
+    None,
+)
+OUT = HERE / "pdf_text.json"
+
+ROMAN_RE = re.compile(r"^[ivxlcdmIVXLCDM]+$")
+# Match a page number on its own line at top/bottom of the page.
+# The book uses Arabic numerals for main chapters (e.g., "3") and Roman for front matter.
+PAGE_NUM_LINE_RE = re.compile(r"^\s*(\d{1,4})\s*$", re.MULTILINE)
+
+
+def detect_book_page(text: str) -> "int | None":
+    """Find the printed page number from standalone page-number lines at the
+    top or bottom of a page."""
+    lines = [l.strip() for l in text.splitlines() if l.strip()]
+    # Check first 2 lines and last 2 lines
+    for candidate in lines[:2] + lines[-2:]:
+        m = re.match(r"^(\d{1,4})$", candidate)
+        if m:
+            return int(m.group(1))
+    return None
+
+
+def main() -> None:
+    if PDF is None:
+        print("No PDF found in project root")
+        return
+
+    print(f"Reading {PDF.name}")
+    reader = pypdf.PdfReader(str(PDF))
+    pages = reader.pages
+    print(f"PDF has {len(pages)} pages")
+
+    by_book_page: dict = {}
+    unmapped: list = []
+    last_seen: "int | None" = None
+    missed_count = 0
+
+    for i, page in enumerate(pages):
+        text = page.extract_text() or ""
+        book_page = detect_book_page(text)
+
+        if book_page is None:
+            # Carry forward sequence: if we saw page N last, assume N+1.
+            if last_seen is not None:
+                book_page = last_seen + 1
+                missed_count += 1
+            else:
+                unmapped.append(i)
+                continue
+        last_seen = book_page
+        # Strip the detected page number from text to clean the output
+        cleaned = re.sub(r"(?m)^\s*\d{1,4}\s*$", "", text).strip()
+        by_book_page[str(book_page)] = {
+            "text": cleaned,
+            "pdfIndex": i,
+        }
+
+    out = {
+        "pdfPageCount": len(pages),
+        "bookPages": by_book_page,
+        "unmapped": unmapped,
+        "inferredPages": missed_count,
+    }
+    OUT.write_text(json.dumps(out, ensure_ascii=False))
+    print(f"Mapped {len(by_book_page)} book pages; {missed_count} inferred; {len(unmapped)} unmapped")
+    print(f"Wrote {OUT}")
+
+
+if __name__ == "__main__":
+    main()

Conjuga/Scripts/textbook/fix_vocab.py

@@ -0,0 +1,252 @@
+#!/usr/bin/env python3
+"""Apply high-confidence auto-fixes from vocab_validation.json to vocab_cards.json.
+
+Auto-fix rules (conservative):
+  1. If a flagged word has exactly one suggestion AND that suggestion differs by
+     <= 2 characters AND has the same starting letter (high-confidence character swap).
+  2. If a card is detected as reversed (Spanish on EN side, English on ES side),
+     swap front/back.
+
+Cards that aren't auto-fixable end up in manual_review.json.
+"""
+
+import json
+import re
+import unicodedata
+from pathlib import Path
+
+HERE = Path(__file__).resolve().parent
+VOCAB = HERE / "vocab_cards.json"
+VALIDATION = HERE / "vocab_validation.json"
+OUT_VOCAB = HERE / "vocab_cards.json"
+OUT_REVIEW = HERE / "manual_review.json"
+OUT_QUARANTINE = HERE / "quarantined_cards.json"
+
+
+def _strip_accents(s: str) -> str:
+    return "".join(c for c in unicodedata.normalize("NFD", s) if unicodedata.category(c) != "Mn")
+
+
+def _levenshtein(a: str, b: str) -> int:
+    if a == b: return 0
+    if not a: return len(b)
+    if not b: return len(a)
+    prev = list(range(len(b) + 1))
+    for i, ca in enumerate(a, 1):
+        curr = [i]
+        for j, cb in enumerate(b, 1):
+            cost = 0 if ca == cb else 1
+            curr.append(min(prev[j] + 1, curr[j - 1] + 1, prev[j - 1] + cost))
+        prev = curr
+    return prev[-1]
+
+
+SPANISH_ACCENT_RE = re.compile(r"[áéíóúñüÁÉÍÓÚÑÜ¿¡]")
+SPANISH_ARTICLES = {"el", "la", "los", "las", "un", "una", "unos", "unas"}
+ENGLISH_STARTERS = {"the", "a", "an", "to", "my", "his", "her", "our", "their"}
+
+
+def language_score(s: str) -> "tuple[int, int]":
+    """Return (es_score, en_score) for a string."""
+    es = 0
+    en = 0
+    if SPANISH_ACCENT_RE.search(s):
+        es += 3
+    words = s.lower().split()
+    if not words:
+        return (es, en)
+    first = words[0].strip(",.;:")
+    if first in SPANISH_ARTICLES:
+        es += 2
+    if first in ENGLISH_STARTERS:
+        en += 2
+    # Spanish-likely endings on later words
+    for w in words:
+        w = w.strip(",.;:")
+        if not w: continue
+        if w.endswith(("ción", "sión", "dad", "tud")):
+            es += 1
+        if w.endswith(("ing", "tion", "ness", "ment", "able", "ly")):
+            en += 1
+    return (es, en)
+
+
+def is_reversed(front: str, back: str) -> bool:
+    """True when front looks like English and back looks like Spanish (i.e. swapped)."""
+    fes, fen = language_score(front)
+    bes, ben = language_score(back)
+    # Front English-leaning AND back Spanish-leaning
+    return fen > fes and bes > ben
+
+
+def best_replacement(word: str, suggestions: list) -> "str | None":
+    """Pick the one safe correction, or None to leave it alone."""
+    if not suggestions:
+        return None
+    # Prefer suggestions that share the same first letter
+    same_initial = [s for s in suggestions if s and word and s[0].lower() == word[0].lower()]
+    candidates = same_initial or suggestions
+    # Single best: short edit distance
+    best = None
+    best_d = 99
+    for s in candidates:
+        d = _levenshtein(word.lower(), s.lower())
+        # Don't apply if the "fix" changes too much
+        if d == 0:
+            continue
+        if d > 2:
+            continue
+        if d < best_d:
+            best = s
+            best_d = d
+    return best
+
+
+def side_language_match(text: str, expected_side: str) -> bool:
+    """Return True when `text` looks like the expected language (es/en).
+    Guards against applying Spanish spell-fix to English words on a mis-paired card.
+    """
+    es, en = language_score(text)
+    if expected_side == "es":
+        return es > en  # require clear Spanish signal
+    if expected_side == "en":
+        return en >= es  # allow equal when text has no strong signal (common for English)
+    return False
+
+
+def apply_word_fixes(text: str, bad_words: list, expected_side: str) -> "tuple[str, list]":
+    """Apply word-level corrections inside a string. Skips fixes entirely when
+    the side's actual language doesn't match the dictionary used, to avoid
+    corrupting mis-paired cards."""
+    if not side_language_match(text, expected_side):
+        return (text, [])
+
+    new_text = text
+    applied = []
+    for bw in bad_words:
+        word = bw["word"]
+        sugg = bw["suggestions"]
+        replacement = best_replacement(word, sugg)
+        if replacement is None:
+            continue
+        # Match standalone word including the (possibly-omitted) trailing period:
+        # `Uds` in the text should be replaced with `Uds.` even when adjacent to `.`.
+        escaped = re.escape(word)
+        # Allow an optional existing period that we'd otherwise duplicate.
+        pattern = re.compile(rf"(?<![A-Za-zÁ-ú]){escaped}\.?(?![A-Za-zÁ-ú])")
+        if pattern.search(new_text):
+            new_text = pattern.sub(replacement, new_text, count=1)
+            applied.append({"from": word, "to": replacement})
+    return (new_text, applied)
+
+
+def main() -> None:
+    vocab_data = json.loads(VOCAB.read_text(encoding="utf-8"))
+    val_data = json.loads(VALIDATION.read_text(encoding="utf-8"))
+
+    # Index validation by (chapter, front, back, sourceImage) for lookup
+    val_index: dict = {}
+    for f in val_data["flags"]:
+        key = (f["chapter"], f["front"], f["back"], f["sourceImage"])
+        val_index[key] = f
+
+    # Walk the cards in place
+    auto_fixed_word = 0
+    auto_swapped = 0
+    quarantined = 0
+    manual_review_cards = []
+    quarantined_cards = []
+
+    for ch in vocab_data["chapters"]:
+        kept_cards = []
+        for card in ch["cards"]:
+            key = (ch["chapter"], card["front"], card["back"], card.get("sourceImage", ""))
+            flag = val_index.get(key)
+
+            # 1) Reversal swap (apply even when not flagged)
+            if is_reversed(card["front"], card["back"]):
+                card["front"], card["back"] = card["back"], card["front"]
+                auto_swapped += 1
+                # Re-key for any further validation lookup (no-op here)
+
+            if flag is None:
+                kept_cards.append(card)
+                continue
+
+            # Quarantine only clear mis-pairs: both sides EXPLICITLY the wrong
+            # language (both Spanish or both English). "unknown" sides stay —
+            # the bounding-box pipeline already handled orientation correctly
+            # and many valid pairs lack the article/accent markers we classify on.
+            fes, fen = language_score(card["front"])
+            bes, ben = language_score(card["back"])
+            front_lang = "es" if fes > fen else ("en" if fen > fes else "unknown")
+            back_lang = "es" if bes > ben else ("en" if ben > bes else "unknown")
+            bothSameLang = (front_lang == "es" and back_lang == "es") or (front_lang == "en" and back_lang == "en")
+            reversed_pair = front_lang == "en" and back_lang == "es"
+            if bothSameLang or reversed_pair:
+                quarantined_cards.append({
+                    "chapter": ch["chapter"],
+                    "front": card["front"],
+                    "back": card["back"],
+                    "sourceImage": card.get("sourceImage", ""),
+                    "reason": f"language-mismatch front={front_lang} back={back_lang}",
+                })
+                quarantined += 1
+                continue
+
+            # 2) Word-level fixes (language-aware)
+            new_front, applied_front = apply_word_fixes(card["front"], flag["badFront"], "es")
+            new_back, applied_back = apply_word_fixes(card["back"], flag["badBack"], "en")
+            card["front"] = new_front
+            card["back"] = new_back
+            auto_fixed_word += len(applied_front) + len(applied_back)
+
+            # If after auto-fix there are STILL flagged words with no
+            # confident replacement, flag for manual review.
+            unresolved_front = [
+                bw for bw in flag["badFront"]
+                if not any(a["from"] == bw["word"] for a in applied_front)
+                and best_replacement(bw["word"], bw["suggestions"]) is None
+            ]
+            unresolved_back = [
+                bw for bw in flag["badBack"]
+                if not any(a["from"] == bw["word"] for a in applied_back)
+                and best_replacement(bw["word"], bw["suggestions"]) is None
+            ]
+            if unresolved_front or unresolved_back:
+                manual_review_cards.append({
+                    "chapter": ch["chapter"],
+                    "front": card["front"],
+                    "back": card["back"],
+                    "sourceImage": card.get("sourceImage", ""),
+                    "unresolvedFront": unresolved_front,
+                    "unresolvedBack": unresolved_back,
+                })
+            kept_cards.append(card)
+
+        ch["cards"] = kept_cards
+
+    OUT_VOCAB.write_text(json.dumps(vocab_data, ensure_ascii=False, indent=2))
+    OUT_REVIEW.write_text(json.dumps({
+        "totalManualReview": len(manual_review_cards),
+        "cards": manual_review_cards,
+    }, ensure_ascii=False, indent=2))
+
+    OUT_QUARANTINE.write_text(json.dumps({
+        "totalQuarantined": len(quarantined_cards),
+        "cards": quarantined_cards,
+    }, ensure_ascii=False, indent=2))
+
+    total_cards = sum(len(c["cards"]) for c in vocab_data["chapters"])
+    print(f"Active cards (after quarantine): {total_cards}")
+    print(f"Auto-swapped (reversed):         {auto_swapped}")
+    print(f"Auto-fixed words:                {auto_fixed_word}")
+    print(f"Quarantined (mis-paired):        {quarantined}")
+    print(f"Cards needing manual review:     {len(manual_review_cards)}")
+    print(f"Wrote {OUT_VOCAB}")
+    print(f"Wrote {OUT_REVIEW}")
+    print(f"Wrote {OUT_QUARANTINE}")
+
+
+if __name__ == "__main__":
+    main()

Conjuga/Scripts/textbook/integrate_repaired.py

@@ -0,0 +1,147 @@
+#!/usr/bin/env python3
+"""Merge repaired_cards.json into vocab_cards.json.
+
+Rules:
+  1. New pairs are added to their chapter's deck if they don't duplicate an existing pair.
+  2. Duplicate detection uses normalize(front)+normalize(back).
+  3. Pairs whose back side starts with a Spanish-article or front side starts
+     with an English article are dropped (pairer got orientation wrong).
+  4. Emits integrate_report.json with counts.
+"""
+
+import json
+import re
+import unicodedata
+from pathlib import Path
+
+HERE = Path(__file__).resolve().parent
+VOCAB = HERE / "vocab_cards.json"
+REPAIRED = HERE / "repaired_cards.json"
+QUARANTINED = HERE / "quarantined_cards.json"
+OUT = HERE / "vocab_cards.json"
+REPORT = HERE / "integrate_report.json"
+
+
+def _strip_accents(s: str) -> str:
+    return "".join(c for c in unicodedata.normalize("NFD", s) if unicodedata.category(c) != "Mn")
+
+
+def norm(s: str) -> str:
+    return _strip_accents(s.lower()).strip()
+
+
+SPANISH_ACCENT_RE = re.compile(r"[áéíóúñüÁÉÍÓÚÑÜ¿¡]")
+SPANISH_ARTICLES = {"el", "la", "los", "las", "un", "una", "unos", "unas"}
+ENGLISH_STARTERS = {"the", "a", "an", "to", "my", "his", "her", "our", "their"}
+
+
+def looks_swapped(front: str, back: str) -> bool:
+    """True if front looks English and back looks Spanish (pair should be swapped)."""
+    fl = front.lower().split()
+    bl = back.lower().split()
+    if not fl or not bl:
+        return False
+    f_first = fl[0].strip(",.;:")
+    b_first = bl[0].strip(",.;:")
+    front_is_en = f_first in ENGLISH_STARTERS
+    back_is_es = (
+        SPANISH_ACCENT_RE.search(back) is not None
+        or b_first in SPANISH_ARTICLES
+    )
+    return front_is_en and back_is_es
+
+
+def looks_good(pair: dict) -> bool:
+    """Basic sanity filter on a repaired pair before it enters the deck."""
+    es = pair["es"].strip()
+    en = pair["en"].strip()
+    if not es or not en: return False
+    if len(es) < 2 or len(en) < 2: return False
+    # Drop if both sides obviously same language (neither has clear orientation)
+    es_has_accent = SPANISH_ACCENT_RE.search(es) is not None
+    en_has_accent = SPANISH_ACCENT_RE.search(en) is not None
+    if en_has_accent and not es_has_accent:
+        # The "en" side has accents — likely swapped
+        return False
+    return True
+
+
+def main() -> None:
+    vocab = json.loads(VOCAB.read_text(encoding="utf-8"))
+    repaired = json.loads(REPAIRED.read_text(encoding="utf-8"))
+    quarantined = json.loads(QUARANTINED.read_text(encoding="utf-8"))
+
+    # Map image → chapter (from the quarantine list — all images here belong to the
+    # chapter they were quarantined from).
+    image_chapter: dict = {}
+    for c in quarantined["cards"]:
+        image_chapter[c["sourceImage"]] = c["chapter"]
+
+    # Build existing key set
+    existing_keys = set()
+    chapter_map: dict = {c["chapter"]: c for c in vocab["chapters"]}
+    for c in vocab["chapters"]:
+        for card in c["cards"]:
+            existing_keys.add((c["chapter"], norm(card["front"]), norm(card["back"])))
+
+    added_per_image: dict = {}
+    dropped_swapped = 0
+    dropped_sanity = 0
+    dropped_dup = 0
+
+    for image_name, data in repaired["byImage"].items():
+        ch_num = image_chapter.get(image_name)
+        if ch_num is None:
+            # Image not in quarantine list (shouldn't happen, but bail)
+            continue
+        deck = chapter_map.setdefault(ch_num, {"chapter": ch_num, "cards": []})
+        added = 0
+        for p in data.get("pairs", []):
+            es = p["es"].strip()
+            en = p["en"].strip()
+            if looks_swapped(es, en):
+                es, en = en, es
+            pair = {"es": es, "en": en}
+            if not looks_good(pair):
+                dropped_sanity += 1
+                continue
+            key = (ch_num, norm(pair["es"]), norm(pair["en"]))
+            if key in existing_keys:
+                dropped_dup += 1
+                continue
+            existing_keys.add(key)
+            card = {
+                "front": pair["es"],
+                "back": pair["en"],
+                "chapter": ch_num,
+                "chapterTitle": "",
+                "section": "",
+                "sourceImage": image_name,
+            }
+            deck["cards"].append(card)
+            added += 1
+        if added:
+            added_per_image[image_name] = added
+
+    # If any new chapter was created, ensure ordered insertion
+    vocab["chapters"] = sorted(chapter_map.values(), key=lambda c: c["chapter"])
+    OUT.write_text(json.dumps(vocab, ensure_ascii=False, indent=2))
+
+    total_added = sum(added_per_image.values())
+    report = {
+        "totalRepairedInput": repaired["totalPairs"],
+        "added": total_added,
+        "dropped_duplicate": dropped_dup,
+        "dropped_sanity": dropped_sanity,
+        "addedPerImage": added_per_image,
+    }
+    REPORT.write_text(json.dumps(report, ensure_ascii=False, indent=2))
+    print(f"Repaired pairs in:          {repaired['totalPairs']}")
+    print(f"Added to deck:              {total_added}")
+    print(f"Dropped as duplicate:       {dropped_dup}")
+    print(f"Dropped as swapped/bad:     {dropped_sanity}")
+    print(f"Wrote {OUT}")
+
+
+if __name__ == "__main__":
+    main()

Conjuga/Scripts/textbook/merge_pdf_into_book.py

@@ -0,0 +1,435 @@
+#!/usr/bin/env python3
+"""Second-pass extractor: use PDF OCR (from ocr_pdf.swift) as a supplementary
+source of clean text, then re-build book.json with PDF-derived content where it
+improves on the EPUB's image-based extraction.
+
+Inputs:
+  chapters.json  — EPUB structural extraction (narrative text + exercise prompts + image refs)
+  answers.json   — EPUB answer key
+  ocr.json       — EPUB image OCR (first pass)
+  pdf_ocr.json   — PDF page-level OCR (this pass, higher DPI + cleaner)
+
+Outputs:
+  book.json         — merged book used by the app
+  vocab_cards.json  — derived vocabulary flashcards
+"""
+
+import json
+import re
+import sys
+from pathlib import Path
+
+HERE = Path(__file__).resolve().parent
+sys.path.insert(0, str(HERE))
+from build_book import (  # reuse the helpers defined in build_book.py
+    COURSE_NAME,
+    build_vocab_cards_for_block,
+    clean_instruction,
+    classify_line,
+    load,
+)
+
+CHAPTERS_JSON = HERE / "chapters.json"
+ANSWERS_JSON = HERE / "answers.json"
+OCR_JSON = HERE / "ocr.json"
+PDF_OCR_JSON = HERE / "pdf_ocr.json"
+PAIRED_VOCAB_JSON = HERE / "paired_vocab.json"  # bounding-box pairs (preferred)
+OUT_BOOK = HERE / "book.json"
+OUT_VOCAB = HERE / "vocab_cards.json"
+
+IMAGE_NAME_RE = re.compile(r"^f(\d{4})-(\d{2})\.jpg$")
+
+
+def extract_book_page(image_src: str) -> "int | None":
+    m = IMAGE_NAME_RE.match(image_src)
+    return int(m.group(1)) if m else None
+
+
+def build_pdf_page_index(pdf_ocr: dict) -> "dict[int, dict]":
+    """Map bookPage → {lines, confidence, pdfIndex}.
+
+    Strategy: use chapter-start alignments as anchors. For each chapter N,
+    anchor[N] = (pdf_idx_where_chapter_starts, book_page_where_chapter_starts).
+    Between anchors we interpolate page-by-page (pages run sequentially within
+    a chapter in this textbook's layout).
+    """
+    pages: "dict[int, dict]" = {}
+    sorted_keys = sorted(pdf_ocr.keys(), key=lambda k: int(k))
+
+    # --- Detect chapter starts in the PDF OCR ---
+    pdf_ch_start: "dict[int, int]" = {}
+    for k in sorted_keys:
+        entry = pdf_ocr[k]
+        lines = entry.get("lines", [])
+        if len(lines) < 2:
+            continue
+        first = lines[0].strip()
+        second = lines[1].strip()
+        if first.isdigit() and 1 <= int(first) <= 30 and len(second) > 5 and second[0:1].isupper():
+            ch = int(first)
+            if ch not in pdf_ch_start:
+                pdf_ch_start[ch] = int(k)
+
+    # --- Load EPUB's authoritative book-page starts ---
+    import re as _re
+    from bs4 import BeautifulSoup as _BS
+    epub_root = HERE.parents[2] / "epub_extract" / "OEBPS"
+    book_ch_start: "dict[int, int]" = {}
+    for ch in sorted(pdf_ch_start.keys()):
+        p = epub_root / f"ch{ch}.xhtml"
+        if not p.exists():
+            continue
+        soup = _BS(p.read_text(encoding="utf-8"), "lxml")
+        for span in soup.find_all(True):
+            id_ = span.get("id", "") or ""
+            m = _re.match(r"page_(\d+)$", id_)
+            if m:
+                book_ch_start[ch] = int(m.group(1))
+                break
+
+    # Build per-chapter (pdf_anchor, book_anchor, next_pdf_anchor) intervals
+    anchors = []  # list of (ch, pdf_start, book_start)
+    for ch in sorted(pdf_ch_start.keys()):
+        if ch in book_ch_start:
+            anchors.append((ch, pdf_ch_start[ch], book_ch_start[ch]))
+
+    for i, (ch, pdf_s, book_s) in enumerate(anchors):
+        next_pdf = anchors[i + 1][1] if i + 1 < len(anchors) else pdf_s + 50
+        # Interpolate book page for each pdf index in [pdf_s, next_pdf)
+        for pdf_idx in range(pdf_s, next_pdf):
+            book_page = book_s + (pdf_idx - pdf_s)
+            entry = pdf_ocr.get(str(pdf_idx))
+            if entry is None:
+                continue
+            if book_page in pages:
+                continue
+            pages[book_page] = {
+                "lines": entry["lines"],
+                "confidence": entry.get("confidence", 0),
+                "pdfIndex": pdf_idx,
+            }
+    return pages
+
+
+def merge_ocr(epub_lines: list, pdf_lines: list) -> list:
+    """EPUB per-image OCR is our primary (targeted, no prose bleed). PDF
+    page-level OCR is only used when EPUB is missing. Per-line accent repair
+    is handled separately via `repair_accents_from_pdf`.
+    """
+    if epub_lines:
+        return epub_lines
+    return pdf_lines
+
+
+import unicodedata as _u
+
+def _strip_accents(s: str) -> str:
+    return "".join(c for c in _u.normalize("NFD", s) if _u.category(c) != "Mn")
+
+
+def _levenshtein(a: str, b: str) -> int:
+    if a == b: return 0
+    if not a: return len(b)
+    if not b: return len(a)
+    prev = list(range(len(b) + 1))
+    for i, ca in enumerate(a, 1):
+        curr = [i]
+        for j, cb in enumerate(b, 1):
+            cost = 0 if ca == cb else 1
+            curr.append(min(prev[j] + 1, curr[j - 1] + 1, prev[j - 1] + cost))
+        prev = curr
+    return prev[-1]
+
+
+def repair_accents_from_pdf(epub_lines: list, pdf_page_lines: list) -> "tuple[list, int]":
+    """For each EPUB OCR line, find a near-match in the PDF page OCR and
+    prefer the PDF version. Repairs include:
+      1. exact accent/case differences (e.g. 'iglesia' vs 'Iglesia')
+      2. single-character OCR errors (e.g. 'the hrother' -> 'the brother')
+      3. two-character OCR errors when the target is long enough
+    """
+    if not epub_lines or not pdf_page_lines:
+        return (epub_lines, 0)
+    # Pre-normalize PDF lines for matching
+    pdf_cleaned = [p.strip() for p in pdf_page_lines if p.strip()]
+    pdf_by_stripped: dict = {}
+    for p in pdf_cleaned:
+        key = _strip_accents(p.lower())
+        pdf_by_stripped.setdefault(key, p)
+
+    out: list = []
+    repairs = 0
+    for e in epub_lines:
+        e_stripped = e.strip()
+        e_key = _strip_accents(e_stripped.lower())
+        # Pass 1: exact accent-only difference
+        if e_key and e_key in pdf_by_stripped and pdf_by_stripped[e_key] != e_stripped:
+            out.append(pdf_by_stripped[e_key])
+            repairs += 1
+            continue
+        # Pass 2: fuzzy — find best PDF line within edit distance 1 or 2
+        if len(e_key) >= 4:
+            max_distance = 1 if len(e_key) < 10 else 2
+            best_match = None
+            best_d = max_distance + 1
+            for p in pdf_cleaned:
+                p_key = _strip_accents(p.lower())
+                # Only match lines of similar length
+                if abs(len(p_key) - len(e_key)) > max_distance:
+                    continue
+                d = _levenshtein(e_key, p_key)
+                if d < best_d:
+                    best_d = d
+                    best_match = p
+                    if d == 0:
+                        break
+            if best_match and best_match != e_stripped and best_d <= max_distance:
+                out.append(best_match)
+                repairs += 1
+                continue
+        out.append(e)
+    return (out, repairs)
+
+
+def vocab_lines_from_pdf_page(
+    pdf_page_entry: dict,
+    epub_narrative_lines: set
+) -> list:
+    """Extract likely vocab-table lines from a PDF page's OCR by filtering out
+    narrative-looking lines (long sentences) and already-known EPUB content."""
+    lines = pdf_page_entry.get("lines", [])
+    out: list = []
+    for raw in lines:
+        line = raw.strip()
+        if not line:
+            continue
+        # Skip lines that look like body prose (too long)
+        if len(line) > 80:
+            continue
+        # Skip narrative we already captured in the EPUB
+        if line in epub_narrative_lines:
+            continue
+        # Skip page-number-only lines
+        if re.fullmatch(r"\d{1,4}", line):
+            continue
+        # Skip standalone chapter headers (e.g. "Nouns, Articles, and Adjectives")
+        out.append(line)
+    return out
+
+
+def main() -> None:
+    chapters_data = load(CHAPTERS_JSON)
+    answers = load(ANSWERS_JSON)["answers"]
+    epub_ocr = load(OCR_JSON)
+    pdf_ocr_raw = load(PDF_OCR_JSON) if PDF_OCR_JSON.exists() else {}
+    pdf_pages = build_pdf_page_index(pdf_ocr_raw) if pdf_ocr_raw else {}
+    paired_vocab = load(PAIRED_VOCAB_JSON) if PAIRED_VOCAB_JSON.exists() else {}
+    print(f"Mapped {len(pdf_pages)} PDF pages to book page numbers")
+    print(f"Loaded bounding-box pairs for {len(paired_vocab)} vocab images")
+
+    # Build a global set of EPUB narrative lines (for subtraction when pulling vocab)
+    narrative_set = set()
+    for ch in chapters_data["chapters"]:
+        for b in ch["blocks"]:
+            if b["kind"] == "paragraph" and b.get("text"):
+                narrative_set.add(b["text"].strip())
+
+    book_chapters = []
+    all_vocab_cards = []
+    pdf_hits = 0
+    pdf_misses = 0
+    merged_pages = 0
+
+    for ch in chapters_data["chapters"]:
+        out_blocks = []
+        current_section_title = ch["title"]
+
+        for bi, block in enumerate(ch["blocks"]):
+            k = block["kind"]
+
+            if k == "heading":
+                current_section_title = block["text"]
+                out_blocks.append(block)
+                continue
+
+            if k == "paragraph":
+                out_blocks.append(block)
+                continue
+
+            if k == "key_vocab_header":
+                out_blocks.append(block)
+                continue
+
+            if k == "vocab_image":
+                src = block["src"]
+                epub_entry = epub_ocr.get(src)
+                epub_lines = epub_entry.get("lines", []) if epub_entry else []
+                epub_conf = epub_entry.get("confidence", 0.0) if epub_entry else 0.0
+
+                book_page = extract_book_page(src)
+                pdf_entry = pdf_pages.get(book_page) if book_page else None
+                pdf_lines = pdf_entry["lines"] if pdf_entry else []
+
+                # Primary: EPUB per-image OCR. Supplementary: PDF page OCR
+                # used only for accent/diacritic repair where keys match.
+                if pdf_lines:
+                    pdf_hits += 1
+                else:
+                    pdf_misses += 1
+                repaired_lines, repairs = repair_accents_from_pdf(epub_lines, pdf_lines)
+                merged_lines = repaired_lines if repaired_lines else pdf_lines
+                merged_conf = max(epub_conf, pdf_entry.get("confidence", 0) if pdf_entry else 0.0)
+                if repairs > 0:
+                    merged_pages += 1
+
+                # Prefer bounding-box pairs (from paired_vocab.json) when
+                # present. Fall back to the block-alternation heuristic.
+                bbox = paired_vocab.get(src, {})
+                bbox_pairs = bbox.get("pairs", []) if isinstance(bbox, dict) else []
+                heuristic = build_vocab_cards_for_block(
+                    {"src": src},
+                    {"lines": merged_lines, "confidence": merged_conf},
+                    ch, current_section_title, bi
+                )
+
+                if bbox_pairs:
+                    cards_for_block = [
+                        {"front": p["es"], "back": p["en"]}
+                        for p in bbox_pairs
+                        if p.get("es") and p.get("en")
+                    ]
+                    # Also feed the flashcard deck
+                    for p in bbox_pairs:
+                        if p.get("es") and p.get("en"):
+                            all_vocab_cards.append({
+                                "front": p["es"],
+                                "back": p["en"],
+                                "chapter": ch["number"],
+                                "chapterTitle": ch["title"],
+                                "section": current_section_title,
+                                "sourceImage": src,
+                            })
+                    pair_source = "bbox"
+                else:
+                    cards_for_block = [{"front": c["front"], "back": c["back"]} for c in heuristic]
+                    all_vocab_cards.extend(heuristic)
+                    pair_source = "heuristic"
+
+                out_blocks.append({
+                    "kind": "vocab_table",
+                    "sourceImage": src,
+                    "ocrLines": merged_lines,
+                    "ocrConfidence": merged_conf,
+                    "cardCount": len(cards_for_block),
+                    "cards": cards_for_block,
+                    "columnCount": bbox.get("columnCount", 2) if isinstance(bbox, dict) else 2,
+                    "source": pair_source,
+                    "bookPage": book_page,
+                    "repairs": repairs,
+                })
+                continue
+
+            if k == "exercise":
+                ans = answers.get(block["id"])
+                # EPUB image OCR (if any image refs)
+                image_ocr_lines: list = []
+                for src in block.get("image_refs", []):
+                    ee = epub_ocr.get(src)
+                    if ee:
+                        image_ocr_lines.extend(ee.get("lines", []))
+                    # Add PDF-page OCR for that page if available
+                    bp = extract_book_page(src)
+                    if bp and pdf_pages.get(bp):
+                        # Only add lines not already present from EPUB OCR
+                        pdf_lines = pdf_pages[bp]["lines"]
+                        for line in pdf_lines:
+                            line = line.strip()
+                            if not line or line in image_ocr_lines:
+                                continue
+                            if line in narrative_set:
+                                continue
+                            image_ocr_lines.append(line)
+
+                prompts = [p for p in block.get("prompts", []) if p.strip()]
+                extras = [e for e in block.get("extra", []) if e.strip()]
+                if not prompts and image_ocr_lines:
+                    # Extract numbered lines from OCR
+                    for line in image_ocr_lines:
+                        m = re.match(r"^(\d+)[.)]\s*(.+)", line.strip())
+                        if m:
+                            prompts.append(f"{m.group(1)}. {m.group(2)}")
+
+                sub = ans["subparts"] if ans else []
+                answer_items = []
+                for sp in sub:
+                    for it in sp["items"]:
+                        answer_items.append({
+                            "label": sp["label"],
+                            "number": it["number"],
+                            "answer": it["answer"],
+                            "alternates": it["alternates"],
+                        })
+
+                out_blocks.append({
+                    "kind": "exercise",
+                    "id": block["id"],
+                    "ansAnchor": block.get("ans_anchor", ""),
+                    "instruction": clean_instruction(block.get("instruction", "")),
+                    "extra": extras,
+                    "prompts": prompts,
+                    "ocrLines": image_ocr_lines,
+                    "freeform": ans["freeform"] if ans else False,
+                    "answerItems": answer_items,
+                    "answerRaw": ans["raw"] if ans else "",
+                    "answerSubparts": sub,
+                })
+                continue
+
+            out_blocks.append(block)
+
+        book_chapters.append({
+            "id": ch["id"],
+            "number": ch["number"],
+            "title": ch["title"],
+            "part": ch.get("part"),
+            "blocks": out_blocks,
+        })
+
+    book = {
+        "courseName": COURSE_NAME,
+        "totalChapters": len(book_chapters),
+        "totalExercises": sum(1 for ch in book_chapters for b in ch["blocks"] if b["kind"] == "exercise"),
+        "totalVocabTables": sum(1 for ch in book_chapters for b in ch["blocks"] if b["kind"] == "vocab_table"),
+        "totalVocabCards": len(all_vocab_cards),
+        "parts": chapters_data.get("part_memberships", {}),
+        "chapters": book_chapters,
+        "sources": {
+            "epub_images_ocr": bool(epub_ocr),
+            "pdf_pages_ocr": bool(pdf_ocr_raw),
+            "pdf_pages_mapped": len(pdf_pages),
+        },
+    }
+    OUT_BOOK.write_text(json.dumps(book, ensure_ascii=False))
+
+    vocab_by_chapter: dict = {}
+    for card in all_vocab_cards:
+        vocab_by_chapter.setdefault(card["chapter"], []).append(card)
+    OUT_VOCAB.write_text(json.dumps({
+        "courseName": COURSE_NAME,
+        "chapters": [
+            {"chapter": n, "cards": cs}
+            for n, cs in sorted(vocab_by_chapter.items())
+        ],
+    }, ensure_ascii=False, indent=2))
+
+    print(f"Wrote {OUT_BOOK}")
+    print(f"Wrote {OUT_VOCAB}")
+    print(f"Chapters:               {book['totalChapters']}")
+    print(f"Exercises:              {book['totalExercises']}")
+    print(f"Vocab tables:           {book['totalVocabTables']}")
+    print(f"Vocab cards (derived):  {book['totalVocabCards']}")
+    print(f"PDF hits vs misses:     {pdf_hits} / {pdf_misses}")
+
+
+if __name__ == "__main__":
+    main()

Conjuga/Scripts/textbook/ocr_all_vocab.swift

@@ -0,0 +1,232 @@
+#!/usr/bin/env swift
+// Bounding-box OCR over every vocab image, producing Spanish→English pairs.
+// Much higher accuracy than the flat-OCR block-alternation heuristic because
+// we use each recognized line's position on the page: rows are clustered by
+// Y-coordinate and cells within a row are split by the biggest X gap.
+//
+// Usage: swift ocr_all_vocab.swift <image_list.json> <oebps_dir> <output.json>
+
+import Foundation
+import Vision
+import AppKit
+
+guard CommandLine.arguments.count >= 4 else {
+    print("Usage: swift ocr_all_vocab.swift <image_list.json> <oebps_dir> <output.json>")
+    exit(1)
+}
+
+let imageListURL = URL(fileURLWithPath: CommandLine.arguments[1])
+let oebpsDir = URL(fileURLWithPath: CommandLine.arguments[2])
+let outputURL = URL(fileURLWithPath: CommandLine.arguments[3])
+
+guard let listData = try? Data(contentsOf: imageListURL),
+      let imageNames = try? JSONDecoder().decode([String].self, from: listData) else {
+    print("Could not load image list at \(imageListURL.path)")
+    exit(1)
+}
+print("Processing \(imageNames.count) images...")
+
+struct RecognizedLine {
+    let text: String
+    let cx: Double
+    let cy: Double
+    let confidence: Double
+}
+
+struct Pair: Encodable {
+    var es: String
+    var en: String
+    var confidence: Double
+}
+
+struct ImageResult: Encodable {
+    var pairs: [Pair]
+    var columnCount: Int
+    var strategy: String
+    var lineCount: Int
+}
+
+let spanishAccents = Set<Character>(["á","é","í","ó","ú","ñ","ü","Á","É","Í","Ó","Ú","Ñ","Ü","¿","¡"])
+let spanishArticles: Set<String> = ["el","la","los","las","un","una","unos","unas"]
+let englishStarters: Set<String> = ["the","a","an","to","my","his","her","our","their","your"]
+let englishOnly: Set<String> = ["the","he","she","it","we","they","is","are","was","were","been","have","has","had","will","would"]
+
+func classify(_ s: String) -> String {
+    let lower = s.lowercased()
+    if lower.contains(where: { spanishAccents.contains($0) }) { return "es" }
+    let first = lower.split(separator: " ").first.map(String.init)?.trimmingCharacters(in: .punctuationCharacters) ?? ""
+    if spanishArticles.contains(first) { return "es" }
+    if englishStarters.contains(first) || englishOnly.contains(first) { return "en" }
+    return "?"
+}
+
+func recognize(_ cgImage: CGImage) -> [RecognizedLine] {
+    let handler = VNImageRequestHandler(cgImage: cgImage, options: [:])
+    let req = VNRecognizeTextRequest()
+    req.recognitionLevel = .accurate
+    req.recognitionLanguages = ["es-ES", "es", "en-US"]
+    req.usesLanguageCorrection = true
+    if #available(macOS 13.0, *) { req.automaticallyDetectsLanguage = true }
+    try? handler.perform([req])
+    var out: [RecognizedLine] = []
+    for obs in req.results ?? [] {
+        guard let top = obs.topCandidates(1).first else { continue }
+        let s = top.string.trimmingCharacters(in: .whitespaces)
+        if s.isEmpty { continue }
+        let bb = obs.boundingBox
+        out.append(RecognizedLine(
+            text: s,
+            cx: Double(bb.origin.x + bb.width / 2),
+            cy: Double(1.0 - (bb.origin.y + bb.height / 2)),
+            confidence: Double(top.confidence)
+        ))
+    }
+    return out
+}
+
+/// Split a sorted-by-X line group into cells by finding the largest gap(s).
+/// `desiredCells` = 2 for 2-col, 4 for 2-pair, etc.
+func splitRow(_ lines: [RecognizedLine], into desiredCells: Int) -> [String] {
+    guard lines.count >= desiredCells else {
+        // Merge into fewer cells: just concatenate left-to-right.
+        return [lines.map(\.text).joined(separator: " ")]
+    }
+    let sorted = lines.sorted { $0.cx < $1.cx }
+    // Find (desiredCells - 1) biggest gaps
+    var gaps: [(idx: Int, gap: Double)] = []
+    for i in 1..<sorted.count {
+        gaps.append((i, sorted[i].cx - sorted[i - 1].cx))
+    }
+    let splitAt = gaps.sorted { $0.gap > $1.gap }
+        .prefix(desiredCells - 1)
+        .map(\.idx)
+        .sorted()
+    var cells: [[RecognizedLine]] = []
+    var start = 0
+    for s in splitAt {
+        cells.append(Array(sorted[start..<s]))
+        start = s
+    }
+    cells.append(Array(sorted[start..<sorted.count]))
+    return cells.map { $0.map(\.text).joined(separator: " ").trimmingCharacters(in: .whitespaces) }
+}
+
+/// Cluster lines into rows by Y proximity. Returns rows in top-to-bottom order.
+func groupRows(_ lines: [RecognizedLine], tol: Double = 0.025) -> [[RecognizedLine]] {
+    let sorted = lines.sorted { $0.cy < $1.cy }
+    var rows: [[RecognizedLine]] = []
+    var current: [RecognizedLine] = []
+    for l in sorted {
+        if let last = current.last, abs(l.cy - last.cy) > tol {
+            rows.append(current)
+            current = [l]
+        } else {
+            current.append(l)
+        }
+    }
+    if !current.isEmpty { rows.append(current) }
+    return rows
+}
+
+/// Detect likely column count: look at how many x-cluster peaks exist across all rows.
+/// Clusters X-coords from all lines into buckets of 10% width.
+func detectColumnCount(_ lines: [RecognizedLine]) -> Int {
+    guard !lines.isEmpty else { return 2 }
+    let step = 0.10
+    var buckets = [Int](repeating: 0, count: Int(1.0 / step) + 1)
+    for l in lines {
+        let b = min(max(0, Int(l.cx / step)), buckets.count - 1)
+        buckets[b] += 1
+    }
+    // A peak = a bucket with count > 10% of total lines
+    let threshold = max(2, lines.count / 10)
+    let peaks = buckets.filter { $0 >= threshold }.count
+    // Most tables are 2-col (peaks = 2). Some 4-col (2 ES/EN pairs side by side → peaks = 4).
+    // Roman/decorative layouts may show 1 peak; treat as 2.
+    switch peaks {
+    case 0, 1, 2: return 2
+    case 3: return 3
+    default: return 4
+    }
+}
+
+/// Merge label-less cells into Spanish→English pairs.
+/// `cells` is a row's cells (length = columnCount). For N=2, [es, en]. For N=4,
+/// [es1, en1, es2, en2] (two pairs). For N=3, [es, en_short, en_long] (rare, merge).
+func cellsToPairs(_ cells: [String], columnCount: Int) -> [(String, String)] {
+    switch columnCount {
+    case 2 where cells.count >= 2:
+        return [(cells[0], cells[1])]
+    case 3 where cells.count >= 3:
+        // 3-col source: es | en | en-alternate. Keep all three by merging EN sides.
+        return [(cells[0], [cells[1], cells[2]].joined(separator: " / "))]
+    case 4 where cells.count >= 4:
+        return [(cells[0], cells[1]), (cells[2], cells[3])]
+    default:
+        if cells.count >= 2 { return [(cells[0], cells.dropFirst().joined(separator: " "))] }
+        return []
+    }
+}
+
+/// Swap pair if orientation is backwards (English on left, Spanish on right).
+func orientPair(_ pair: (String, String)) -> (String, String) {
+    let (a, b) = pair
+    let ca = classify(a), cb = classify(b)
+    if ca == "en" && cb == "es" { return (b, a) }
+    return pair
+}
+
+var results: [String: ImageResult] = [:]
+var processed = 0
+let startTime = Date()
+
+for name in imageNames {
+    processed += 1
+    let url = oebpsDir.appendingPathComponent(name)
+    guard let nsImg = NSImage(contentsOf: url),
+          let tiff = nsImg.tiffRepresentation,
+          let rep = NSBitmapImageRep(data: tiff),
+          let cg = rep.cgImage else {
+        continue
+    }
+    let lines = recognize(cg)
+    if lines.isEmpty {
+        results[name] = ImageResult(pairs: [], columnCount: 2, strategy: "empty", lineCount: 0)
+        continue
+    }
+
+    let columnCount = detectColumnCount(lines)
+    let rows = groupRows(lines, tol: 0.025)
+    var pairs: [Pair] = []
+    for row in rows {
+        guard row.count >= 2 else { continue }
+        let cells = splitRow(row, into: columnCount)
+        let rawPairs = cellsToPairs(cells, columnCount: columnCount)
+        for p in rawPairs {
+            let (es, en) = orientPair(p)
+            if es.count < 1 || en.count < 1 { continue }
+            let avgConf = row.reduce(0.0) { $0 + $1.confidence } / Double(row.count)
+            pairs.append(Pair(es: es, en: en, confidence: avgConf))
+        }
+    }
+    results[name] = ImageResult(
+        pairs: pairs,
+        columnCount: columnCount,
+        strategy: "bbox-row-split",
+        lineCount: lines.count
+    )
+
+    if processed % 50 == 0 || processed == imageNames.count {
+        let elapsed = Date().timeIntervalSince(startTime)
+        let rate = Double(processed) / max(elapsed, 0.001)
+        let eta = Double(imageNames.count - processed) / max(rate, 0.001)
+        print(String(format: "%d/%d  %.1f img/s  eta %.0fs", processed, imageNames.count, rate, eta))
+    }
+}
+
+let enc = JSONEncoder()
+enc.outputFormatting = [.sortedKeys]
+try enc.encode(results).write(to: outputURL)
+let totalPairs = results.values.reduce(0) { $0 + $1.pairs.count }
+let emptyTables = results.values.filter { $0.pairs.isEmpty }.count
+print("Wrote \(results.count) results, \(totalPairs) total pairs, \(emptyTables) unpaired")

Conjuga/Scripts/textbook/ocr_images.swift

@@ -0,0 +1,110 @@
+#!/usr/bin/env swift
+// OCR every JPG in the given input directory using the macOS Vision framework.
+// Output: JSON map of { "<filename>": { "lines": [...], "confidence": Double } }
+//
+// Usage: swift ocr_images.swift <input_dir> <output_json>
+// Example: swift ocr_images.swift ../../../epub_extract/OEBPS ocr.json
+
+import Foundation
+import Vision
+import AppKit
+
+guard CommandLine.arguments.count >= 3 else {
+    print("Usage: swift ocr_images.swift <input_dir> <output_json>")
+    exit(1)
+}
+
+let inputDir = URL(fileURLWithPath: CommandLine.arguments[1])
+let outputURL = URL(fileURLWithPath: CommandLine.arguments[2])
+
+// Skip images that are icons/inline markers — not real content
+let skipSubstrings = ["Common", "cover", "title"]
+
+let fileManager = FileManager.default
+guard let enumerator = fileManager.enumerator(at: inputDir, includingPropertiesForKeys: nil) else {
+    print("Could not enumerate \(inputDir.path)")
+    exit(1)
+}
+
+var jpgs: [URL] = []
+for case let url as URL in enumerator {
+    let name = url.lastPathComponent
+    guard name.hasSuffix(".jpg") || name.hasSuffix(".jpeg") || name.hasSuffix(".png") else { continue }
+    if skipSubstrings.contains(where: { name.contains($0) }) { continue }
+    jpgs.append(url)
+}
+jpgs.sort { $0.lastPathComponent < $1.lastPathComponent }
+print("Found \(jpgs.count) images to OCR")
+
+struct OCRResult: Encodable {
+    var lines: [String]
+    var confidence: Double
+}
+
+var results: [String: OCRResult] = [:]
+let total = jpgs.count
+var processed = 0
+let startTime = Date()
+
+for url in jpgs {
+    processed += 1
+    let name = url.lastPathComponent
+
+    guard let nsImage = NSImage(contentsOf: url),
+          let tiffData = nsImage.tiffRepresentation,
+          let bitmap = NSBitmapImageRep(data: tiffData),
+          let cgImage = bitmap.cgImage else {
+        print("\(processed)/\(total) \(name) — could not load")
+        continue
+    }
+
+    let handler = VNImageRequestHandler(cgImage: cgImage, options: [:])
+    let request = VNRecognizeTextRequest()
+    request.recognitionLevel = .accurate
+    request.recognitionLanguages = ["es-ES", "es", "en-US"]
+    request.usesLanguageCorrection = true
+    // For the 2020 book, automaticallyDetectsLanguage helps with mixed content
+    if #available(macOS 13.0, *) {
+        request.automaticallyDetectsLanguage = true
+    }
+
+    do {
+        try handler.perform([request])
+        let observations = request.results ?? []
+        var lines: [String] = []
+        var totalConfidence: Float = 0
+        var count = 0
+        for obs in observations {
+            if let top = obs.topCandidates(1).first {
+                let s = top.string.trimmingCharacters(in: .whitespaces)
+                if !s.isEmpty {
+                    lines.append(s)
+                    totalConfidence += top.confidence
+                    count += 1
+                }
+            }
+        }
+        let avg = count > 0 ? Double(totalConfidence) / Double(count) : 0.0
+        results[name] = OCRResult(lines: lines, confidence: avg)
+    } catch {
+        print("\(processed)/\(total) \(name) — error: \(error)")
+    }
+
+    if processed % 50 == 0 || processed == total {
+        let elapsed = Date().timeIntervalSince(startTime)
+        let rate = Double(processed) / max(elapsed, 0.001)
+        let remaining = Double(total - processed) / max(rate, 0.001)
+        print(String(format: "%d/%d  %.1f img/s  eta %.0fs", processed, total, rate, remaining))
+    }
+}
+
+let encoder = JSONEncoder()
+encoder.outputFormatting = [.prettyPrinted, .sortedKeys]
+do {
+    let data = try encoder.encode(results)
+    try data.write(to: outputURL)
+    print("Wrote \(results.count) OCR entries to \(outputURL.path)")
+} catch {
+    print("Error writing output: \(error)")
+    exit(1)
+}

Conjuga/Scripts/textbook/ocr_pdf.swift

@@ -0,0 +1,133 @@
+#!/usr/bin/env swift
+// Rasterize each page of a PDF at high DPI and OCR it with Vision.
+// Output: { "<pdfIndex>": { "lines": [...], "confidence": Double, "bookPage": Int? } }
+//
+// Usage: swift ocr_pdf.swift <pdf_path> <output_json> [dpi]
+// Example: swift ocr_pdf.swift "book.pdf" pdf_ocr.json 240
+
+import Foundation
+import Vision
+import AppKit
+import Quartz
+
+guard CommandLine.arguments.count >= 3 else {
+    print("Usage: swift ocr_pdf.swift <pdf_path> <output_json> [dpi]")
+    exit(1)
+}
+
+let pdfURL = URL(fileURLWithPath: CommandLine.arguments[1])
+let outputURL = URL(fileURLWithPath: CommandLine.arguments[2])
+let dpi: CGFloat = CommandLine.arguments.count >= 4 ? CGFloat(Double(CommandLine.arguments[3]) ?? 240.0) : 240.0
+
+guard let pdfDoc = PDFDocument(url: pdfURL) else {
+    print("Could not open PDF at \(pdfURL.path)")
+    exit(1)
+}
+
+let pageCount = pdfDoc.pageCount
+print("PDF has \(pageCount) pages. Rendering at \(dpi) DPI.")
+
+struct PageResult: Encodable {
+    var lines: [String]
+    var confidence: Double
+    var bookPage: Int?
+}
+
+var results: [String: PageResult] = [:]
+let startTime = Date()
+
+// Render at scale = dpi / 72 (72 is default PDF DPI)
+let scale: CGFloat = dpi / 72.0
+
+for i in 0..<pageCount {
+    guard let page = pdfDoc.page(at: i) else { continue }
+    let pageBounds = page.bounds(for: .mediaBox)
+    let scaledSize = CGSize(width: pageBounds.width * scale, height: pageBounds.height * scale)
+
+    // Render the page into a CGImage
+    let colorSpace = CGColorSpaceCreateDeviceRGB()
+    let bitmapInfo = CGImageAlphaInfo.noneSkipLast.rawValue
+    guard let context = CGContext(
+        data: nil,
+        width: Int(scaledSize.width),
+        height: Int(scaledSize.height),
+        bitsPerComponent: 8,
+        bytesPerRow: 0,
+        space: colorSpace,
+        bitmapInfo: bitmapInfo
+    ) else {
+        print("\(i): could not create CGContext")
+        continue
+    }
+    context.setFillColor(CGColor(gray: 1.0, alpha: 1.0))
+    context.fill(CGRect(origin: .zero, size: scaledSize))
+    context.scaleBy(x: scale, y: scale)
+    page.draw(with: .mediaBox, to: context)
+
+    guard let cgImage = context.makeImage() else {
+        print("\(i): could not create CGImage")
+        continue
+    }
+
+    let handler = VNImageRequestHandler(cgImage: cgImage, options: [:])
+    let request = VNRecognizeTextRequest()
+    request.recognitionLevel = .accurate
+    request.recognitionLanguages = ["es-ES", "es", "en-US"]
+    request.usesLanguageCorrection = true
+    if #available(macOS 13.0, *) {
+        request.automaticallyDetectsLanguage = true
+    }
+
+    do {
+        try handler.perform([request])
+        let observations = request.results ?? []
+        var lines: [String] = []
+        var totalConfidence: Float = 0
+        var count = 0
+        for obs in observations {
+            if let top = obs.topCandidates(1).first {
+                let s = top.string.trimmingCharacters(in: .whitespaces)
+                if !s.isEmpty {
+                    lines.append(s)
+                    totalConfidence += top.confidence
+                    count += 1
+                }
+            }
+        }
+        let avg = count > 0 ? Double(totalConfidence) / Double(count) : 0.0
+
+        // Try to detect book page number: a short numeric line in the first
+        // 3 or last 3 entries (typical page-number placement).
+        var bookPage: Int? = nil
+        let candidates = Array(lines.prefix(3)) + Array(lines.suffix(3))
+        for c in candidates {
+            let trimmed = c.trimmingCharacters(in: .whitespaces)
+            if let n = Int(trimmed), n >= 1 && n <= 1000 {
+                bookPage = n
+                break
+            }
+        }
+
+        results[String(i)] = PageResult(lines: lines, confidence: avg, bookPage: bookPage)
+    } catch {
+        print("\(i): \(error)")
+    }
+
+    if (i + 1) % 25 == 0 || (i + 1) == pageCount {
+        let elapsed = Date().timeIntervalSince(startTime)
+        let rate = Double(i + 1) / max(elapsed, 0.001)
+        let remaining = Double(pageCount - (i + 1)) / max(rate, 0.001)
+        print(String(format: "%d/%d  %.1f pg/s  eta %.0fs", i + 1, pageCount, rate, remaining))
+    }
+}
+
+let encoder = JSONEncoder()
+encoder.outputFormatting = [.sortedKeys]
+do {
+    let data = try encoder.encode(results)
+    try data.write(to: outputURL)
+    print("Wrote \(results.count) pages to \(outputURL.path)")
+} catch {
+    print("Error writing output: \(error)")
+    exit(1)
+}

Conjuga/Scripts/textbook/repair_quarantined.swift

@@ -0,0 +1,177 @@
+#!/usr/bin/env swift
+// Re-OCR the images referenced in quarantined_cards.json using Vision with
+// bounding-box info, then pair lines by column position (left = Spanish,
+// right = English) instead of by document read order.
+//
+// Output: repaired_cards.json — {"byImage": {"f0142-02.jpg": [{"es":..., "en":...}, ...]}}
+
+import Foundation
+import Vision
+import AppKit
+
+guard CommandLine.arguments.count >= 4 else {
+    print("Usage: swift repair_quarantined.swift <quarantined.json> <epub_oebps_dir> <output.json>")
+    exit(1)
+}
+
+let quarantinedURL = URL(fileURLWithPath: CommandLine.arguments[1])
+let imageDir = URL(fileURLWithPath: CommandLine.arguments[2])
+let outputURL = URL(fileURLWithPath: CommandLine.arguments[3])
+
+guard let data = try? Data(contentsOf: quarantinedURL),
+      let json = try? JSONSerialization.jsonObject(with: data) as? [String: Any],
+      let cards = json["cards"] as? [[String: Any]] else {
+    print("Could not load \(quarantinedURL.path)")
+    exit(1)
+}
+
+var uniqueImages = Set<String>()
+for card in cards {
+    if let src = card["sourceImage"] as? String { uniqueImages.insert(src) }
+}
+print("Unique images to re-OCR: \(uniqueImages.count)")
+
+struct RecognizedLine {
+    let text: String
+    let cx: CGFloat   // center X (normalized 0..1)
+    let cy: CGFloat   // center Y (normalized 0..1 from top)
+    let confidence: Float
+}
+
+struct Pair: Encodable {
+    var es: String
+    var en: String
+    var confidence: Double
+}
+
+struct ImageResult: Encodable {
+    var pairs: [Pair]
+    var lineCount: Int
+    var strategy: String
+}
+
+func classify(_ s: String) -> String {
+    // "es" if has accents or starts with ES article; "en" if starts with EN article; else "?"
+    let lower = s.lowercased()
+    let accentChars: Set<Character> = ["á", "é", "í", "ó", "ú", "ñ", "ü", "¿", "¡"]
+    if lower.contains(where: { accentChars.contains($0) }) { return "es" }
+    let first = lower.split(separator: " ").first.map(String.init) ?? ""
+    let esArticles: Set<String> = ["el", "la", "los", "las", "un", "una", "unos", "unas"]
+    let enStarters: Set<String> = ["the", "a", "an", "to", "my", "his", "her", "our", "their"]
+    if esArticles.contains(first) { return "es" }
+    if enStarters.contains(first) { return "en" }
+    return "?"
+}
+
+func recognizeLines(cgImage: CGImage) -> [RecognizedLine] {
+    let handler = VNImageRequestHandler(cgImage: cgImage, options: [:])
+    let request = VNRecognizeTextRequest()
+    request.recognitionLevel = .accurate
+    request.recognitionLanguages = ["es-ES", "es", "en-US"]
+    request.usesLanguageCorrection = true
+    if #available(macOS 13.0, *) {
+        request.automaticallyDetectsLanguage = true
+    }
+    do { try handler.perform([request]) } catch { return [] }
+    var out: [RecognizedLine] = []
+    for obs in request.results ?? [] {
+        guard let top = obs.topCandidates(1).first else { continue }
+        let s = top.string.trimmingCharacters(in: .whitespaces)
+        if s.isEmpty { continue }
+        // Vision's boundingBox is normalized with origin at lower-left
+        let bb = obs.boundingBox
+        let cx = bb.origin.x + bb.width / 2
+        let cyTop = 1.0 - (bb.origin.y + bb.height / 2)  // flip to top-origin
+        out.append(RecognizedLine(text: s, cx: cx, cy: cyTop, confidence: top.confidence))
+    }
+    return out
+}
+
+/// Pair lines by column position: left column = Spanish, right column = English.
+/// Groups lines into rows by Y proximity, then within each row pairs left-right.
+func pairByPosition(_ lines: [RecognizedLine]) -> ([Pair], String) {
+    guard !lines.isEmpty else { return ([], "empty") }
+
+    // Cluster by Y into rows. Use adaptive row height: median line gap * 0.6
+    let sortedByY = lines.sorted { $0.cy < $1.cy }
+    var rows: [[RecognizedLine]] = []
+    var current: [RecognizedLine] = []
+    let rowTol: CGFloat = 0.015   // 1.5% of page height
+    for l in sortedByY {
+        if let last = current.last, abs(l.cy - last.cy) > rowTol {
+            rows.append(current)
+            current = [l]
+        } else {
+            current.append(l)
+        }
+    }
+    if !current.isEmpty { rows.append(current) }
+
+    var pairs: [Pair] = []
+    var strategy = "row-pair"
+    for row in rows {
+        guard row.count >= 2 else { continue }
+        // Sort row by X, split at midpoint; left = Spanish, right = English
+        let sortedX = row.sorted { $0.cx < $1.cx }
+        // Find gap: pick the biggest x-gap in the row to split
+        var maxGap: CGFloat = 0
+        var splitIdx = 1
+        for i in 1..<sortedX.count {
+            let gap = sortedX[i].cx - sortedX[i - 1].cx
+            if gap > maxGap {
+                maxGap = gap
+                splitIdx = i
+            }
+        }
+        let leftLines = Array(sortedX[0..<splitIdx])
+        let rightLines = Array(sortedX[splitIdx..<sortedX.count])
+        let leftText = leftLines.map(\.text).joined(separator: " ").trimmingCharacters(in: .whitespaces)
+        let rightText = rightLines.map(\.text).joined(separator: " ").trimmingCharacters(in: .whitespaces)
+        if leftText.isEmpty || rightText.isEmpty { continue }
+        // Verify language orientation — swap if we got it backwards
+        var es = leftText
+        var en = rightText
+        let lc = classify(es)
+        let rc = classify(en)
+        if lc == "en" && rc == "es" {
+            es = rightText
+            en = leftText
+        }
+        let avgConf = (leftLines + rightLines).reduce(Float(0)) { $0 + $1.confidence } / Float(leftLines.count + rightLines.count)
+        pairs.append(Pair(es: es, en: en, confidence: Double(avgConf)))
+    }
+
+    if pairs.isEmpty { strategy = "no-rows" }
+    return (pairs, strategy)
+}
+
+var results: [String: ImageResult] = [:]
+
+for name in uniqueImages.sorted() {
+    let url = imageDir.appendingPathComponent(name)
+    guard let img = NSImage(contentsOf: url),
+          let tiff = img.tiffRepresentation,
+          let rep = NSBitmapImageRep(data: tiff),
+          let cg = rep.cgImage else {
+        print("\(name): could not load")
+        continue
+    }
+    let lines = recognizeLines(cgImage: cg)
+    let (pairs, strategy) = pairByPosition(lines)
+    results[name] = ImageResult(pairs: pairs, lineCount: lines.count, strategy: strategy)
+    print("\(name): \(lines.count) lines -> \(pairs.count) pairs via \(strategy)")
+}
+
+struct Output: Encodable {
+    var byImage: [String: ImageResult]
+    var totalPairs: Int
+}
+let output = Output(
+    byImage: results,
+    totalPairs: results.values.reduce(0) { $0 + $1.pairs.count }
+)
+
+let enc = JSONEncoder()
+enc.outputFormatting = [.prettyPrinted, .sortedKeys]
+try enc.encode(output).write(to: outputURL)
+print("Wrote \(output.totalPairs) repaired pairs to \(outputURL.path)")

Conjuga/Scripts/textbook/run_pipeline.sh

@@ -0,0 +1,54 @@
+#!/usr/bin/env bash
+# End-to-end textbook extraction pipeline.
+#
+# Requires: Python 3 + lxml/beautifulsoup4/pypdf installed.
+#           macOS for Vision + NSSpellChecker (Swift).
+#
+# Inputs: EPUB extracted to epub_extract/OEBPS/ and the PDF at project root.
+# Outputs: book.json, vocab_cards.json, manual_review.json, quarantined_cards.json
+
+set -e
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+ROOT="$(cd "$SCRIPT_DIR/../../.." && pwd)"
+cd "$ROOT"
+
+echo "=== Phase 1a: parse XHTML chapters ==="
+python3 "$SCRIPT_DIR/extract_chapters.py"
+
+echo "=== Phase 1b: parse answer key ==="
+python3 "$SCRIPT_DIR/extract_answers.py"
+
+if [ ! -f "$SCRIPT_DIR/ocr.json" ]; then
+    echo "=== Phase 1c: OCR EPUB images (first-time only) ==="
+    swift "$SCRIPT_DIR/ocr_images.swift" "$ROOT/epub_extract/OEBPS" "$SCRIPT_DIR/ocr.json"
+else
+    echo "=== Phase 1c: EPUB OCR already cached ==="
+fi
+
+PDF_FILE="$(ls "$ROOT"/Complete\ Spanish\ Step-By-Step*.pdf 2>/dev/null | head -1 || true)"
+if [ -n "$PDF_FILE" ] && [ ! -f "$SCRIPT_DIR/pdf_ocr.json" ]; then
+    echo "=== Phase 1d: OCR PDF pages (first-time only) ==="
+    swift "$SCRIPT_DIR/ocr_pdf.swift" "$PDF_FILE" "$SCRIPT_DIR/pdf_ocr.json" 240
+fi
+
+echo "=== Phase 1e: merge into book.json ==="
+python3 "$SCRIPT_DIR/merge_pdf_into_book.py"
+
+echo "=== Phase 2: spell-check validation ==="
+swift "$SCRIPT_DIR/validate_vocab.swift" "$SCRIPT_DIR/vocab_cards.json" "$SCRIPT_DIR/vocab_validation.json"
+
+echo "=== Phase 3: auto-fix + quarantine pass 1 ==="
+python3 "$SCRIPT_DIR/fix_vocab.py"
+
+echo "=== Phase 3: auto-fix + quarantine pass 2 (convergence) ==="
+swift "$SCRIPT_DIR/validate_vocab.swift" "$SCRIPT_DIR/vocab_cards.json" "$SCRIPT_DIR/vocab_validation.json"
+python3 "$SCRIPT_DIR/fix_vocab.py"
+
+echo ""
+echo "=== Copy to app bundle ==="
+cp "$SCRIPT_DIR/book.json" "$ROOT/Conjuga/Conjuga/textbook_data.json"
+cp "$SCRIPT_DIR/vocab_cards.json" "$ROOT/Conjuga/Conjuga/textbook_vocab.json"
+ls -lh "$ROOT/Conjuga/Conjuga/textbook_"*.json
+echo ""
+echo "Done. Bump textbookDataVersion in DataLoader.swift to trigger re-seed."

Conjuga/Scripts/textbook/validate_vocab.swift

@@ -0,0 +1,156 @@
+#!/usr/bin/env swift
+// Validate every Spanish/English word in vocab_cards.json using NSSpellChecker.
+// For each flagged word, produce up to 3 candidate corrections.
+//
+// Usage: swift validate_vocab.swift <vocab_cards.json> <output_report.json>
+
+import Foundation
+import AppKit
+
+guard CommandLine.arguments.count >= 3 else {
+    print("Usage: swift validate_vocab.swift <vocab_cards.json> <output_report.json>")
+    exit(1)
+}
+
+let inputURL = URL(fileURLWithPath: CommandLine.arguments[1])
+let outputURL = URL(fileURLWithPath: CommandLine.arguments[2])
+
+guard let data = try? Data(contentsOf: inputURL),
+      let json = try? JSONSerialization.jsonObject(with: data) as? [String: Any],
+      let chapters = json["chapters"] as? [[String: Any]] else {
+    print("Could not load \(inputURL.path)")
+    exit(1)
+}
+
+let checker = NSSpellChecker.shared
+
+// Tokenize — only letter runs (Unicode aware for Spanish accents)
+func tokens(_ s: String) -> [String] {
+    let letters = CharacterSet.letters
+    return s.unicodeScalars
+        .split { !letters.contains($0) }
+        .map { String(String.UnicodeScalarView($0)) }
+        .filter { !$0.isEmpty }
+}
+
+// Minimal stopword set — names, proper nouns, numeric tokens already filtered
+let stopES: Set<String> = [
+    "el", "la", "los", "las", "un", "una", "unos", "unas", "del", "al", "de",
+    "a", "en", "y", "o", "que", "no", "se", "con", "por", "para", "lo", "le",
+    "su", "mi", "tu", "yo", "te", "me", "es", "son", "está", "están",
+]
+let stopEN: Set<String> = [
+    "the", "a", "an", "to", "of", "in", "and", "or", "is", "are", "was", "were",
+    "be", "been", "my", "his", "her", "our", "their", "your",
+]
+
+func checkWord(_ w: String, lang: String, stop: Set<String>) -> [String]? {
+    // Return nil if word is OK, else list of candidate corrections.
+    if w.count < 2 { return nil }
+    if stop.contains(w.lowercased()) { return nil }
+    if w.rangeOfCharacter(from: .decimalDigits) != nil { return nil }
+
+    let range = checker.checkSpelling(
+        of: w,
+        startingAt: 0,
+        language: lang,
+        wrap: false,
+        inSpellDocumentWithTag: 0,
+        wordCount: nil
+    )
+    // Range of `(0, 0)` means no misspelling; otherwise we have a misspelling.
+    if range.location == NSNotFound || range.length == 0 { return nil }
+
+    let guesses = checker.guesses(
+        forWordRange: NSRange(location: 0, length: (w as NSString).length),
+        in: w,
+        language: lang,
+        inSpellDocumentWithTag: 0
+    ) ?? []
+    return Array(guesses.prefix(3))
+}
+
+struct Flag: Encodable {
+    var chapter: Int
+    var front: String
+    var back: String
+    var badFront: [BadWord]
+    var badBack: [BadWord]
+    var sourceImage: String
+}
+struct BadWord: Encodable {
+    var word: String
+    var suggestions: [String]
+    var side: String  // "es" or "en"
+}
+
+var flags: [Flag] = []
+var totalCards = 0
+var totalBadES = 0
+var totalBadEN = 0
+
+for ch in chapters {
+    guard let chNum = ch["chapter"] as? Int,
+          let cards = ch["cards"] as? [[String: Any]] else { continue }
+    for card in cards {
+        totalCards += 1
+        let front = (card["front"] as? String) ?? ""
+        let back = (card["back"] as? String) ?? ""
+        let img = (card["sourceImage"] as? String) ?? ""
+
+        var badFront: [BadWord] = []
+        for w in tokens(front) {
+            if let sugg = checkWord(w, lang: "es", stop: stopES) {
+                badFront.append(BadWord(word: w, suggestions: sugg, side: "es"))
+                totalBadES += 1
+            }
+        }
+        var badBack: [BadWord] = []
+        for w in tokens(back) {
+            if let sugg = checkWord(w, lang: "en", stop: stopEN) {
+                badBack.append(BadWord(word: w, suggestions: sugg, side: "en"))
+                totalBadEN += 1
+            }
+        }
+        if !badFront.isEmpty || !badBack.isEmpty {
+            flags.append(Flag(
+                chapter: chNum,
+                front: front,
+                back: back,
+                badFront: badFront,
+                badBack: badBack,
+                sourceImage: img
+            ))
+        }
+    }
+}
+
+struct Report: Encodable {
+    var totalCards: Int
+    var flaggedCards: Int
+    var flaggedSpanishWords: Int
+    var flaggedEnglishWords: Int
+    var flags: [Flag]
+}
+let report = Report(
+    totalCards: totalCards,
+    flaggedCards: flags.count,
+    flaggedSpanishWords: totalBadES,
+    flaggedEnglishWords: totalBadEN,
+    flags: flags
+)
+
+let encoder = JSONEncoder()
+encoder.outputFormatting = [.prettyPrinted, .sortedKeys]
+do {
+    let data = try encoder.encode(report)
+    try data.write(to: outputURL)
+    print("Cards:              \(totalCards)")
+    print("Flagged cards:      \(flags.count)  (\(Double(flags.count)/Double(totalCards)*100.0 as Double)%)")
+    print("Flagged ES words:   \(totalBadES)")
+    print("Flagged EN words:   \(totalBadEN)")
+    print("Wrote \(outputURL.path)")
+} catch {
+    print("Error writing output: \(error)")
+    exit(1)
+}

Conjuga/SharedModels/Sources/SharedModels/AnswerGrader.swift

@@ -0,0 +1,68 @@
+import Foundation
+
+/// On-device deterministic answer grader with partial-credit support.
+/// No network calls, no API keys. Handles accent stripping and single-char typos.
+public enum AnswerGrader {
+
+    /// Evaluate `userText` against the canonical answer (plus alternates).
+    /// Returns `.correct` for exact/normalized match, `.close` for accent-strip
+    /// match or Levenshtein distance 1, `.wrong` otherwise.
+    public static func grade(userText: String, canonical: String, alternates: [String] = []) -> TextbookGrade {
+        let candidates = [canonical] + alternates
+        let normalizedUser = normalize(userText)
+        if normalizedUser.isEmpty { return .wrong }
+
+        for c in candidates {
+            if normalize(c) == normalizedUser { return .correct }
+        }
+        for c in candidates {
+            if stripAccents(normalize(c)) == stripAccents(normalizedUser) {
+                return .close
+            }
+        }
+        for c in candidates {
+            if levenshtein(normalizedUser, normalize(c)) <= 1 {
+                return .close
+            }
+        }
+        return .wrong
+    }
+
+    /// Lowercase, collapse whitespace, strip leading/trailing punctuation.
+    public static func normalize(_ s: String) -> String {
+        let lowered = s.lowercased(with: Locale(identifier: "es"))
+        let collapsed = lowered.replacingOccurrences(of: "\\s+", with: " ", options: .regularExpression)
+        let trimmed = collapsed.trimmingCharacters(in: .whitespacesAndNewlines)
+        let punct = CharacterSet(charactersIn: ".,;:!?¿¡\"'()[]{}—–-")
+        return trimmed.trimmingCharacters(in: punct)
+    }
+
+    /// Remove combining diacritics (á→a, ñ→n, ü→u).
+    public static func stripAccents(_ s: String) -> String {
+        s.folding(options: .diacriticInsensitive, locale: Locale(identifier: "en"))
+    }
+
+    /// Standard Levenshtein edit distance.
+    public static func levenshtein(_ a: String, _ b: String) -> Int {
+        if a == b { return 0 }
+        if a.isEmpty { return b.count }
+        if b.isEmpty { return a.count }
+        let aa = Array(a)
+        let bb = Array(b)
+        var prev = Array(0...bb.count)
+        var curr = Array(repeating: 0, count: bb.count + 1)
+        for i in 1...aa.count {
+            curr[0] = i
+            for j in 1...bb.count {
+                let cost = aa[i - 1] == bb[j - 1] ? 0 : 1
+                curr[j] = min(
+                    prev[j] + 1,
+                    curr[j - 1] + 1,
+                    prev[j - 1] + cost
+                )
+            }
+            swap(&prev, &curr)
+        }
+        return prev[bb.count]
+    }
+}

Conjuga/SharedModels/Sources/SharedModels/PracticeFilter.swift

@@ -0,0 +1,81 @@
+import Foundation
+
+/// Pure practice-pool filtering (Issue #26).
+///
+/// Takes plain value snapshots of the verb + irregular-span data and computes
+/// the set of verb IDs eligible for practice under the user's selected filters.
+/// Deliberately decoupled from SwiftData so the same logic is directly testable
+/// without a ModelContainer.
+public enum PracticeFilter {
+
+    /// Minimal verb snapshot for filtering.
+    public struct VerbSlot: Sendable, Hashable {
+        public let id: Int
+        public let level: String
+        public init(id: Int, level: String) {
+            self.id = id
+            self.level = level
+        }
+    }
+
+    /// Minimal irregular-span snapshot for filtering.
+    public struct IrregularSlot: Sendable, Hashable {
+        public let verbId: Int
+        public let category: IrregularSpan.SpanCategory
+        public init(verbId: Int, category: IrregularSpan.SpanCategory) {
+            self.verbId = verbId
+            self.category = category
+        }
+    }
+
+    /// Union of `VerbLevelGroup.dataLevels(for:)` across every user-facing level.
+    /// An empty input produces an empty result; callers decide the empty semantics.
+    public static func dataLevels(forSelectedLevels levels: Set<String>) -> Set<String> {
+        levels.reduce(into: Set<String>()) { acc, level in
+            acc.formUnion(VerbLevelGroup.dataLevels(for: level))
+        }
+    }
+
+    /// Verb IDs whose `level` falls inside any of the selected level groups.
+    public static func verbIDs(
+        matchingLevels selectedLevels: Set<String>,
+        in verbs: [VerbSlot]
+    ) -> Set<Int> {
+        guard !selectedLevels.isEmpty else { return [] }
+        let expanded = dataLevels(forSelectedLevels: selectedLevels)
+        return Set(verbs.filter { expanded.contains($0.level) }.map(\.id))
+    }
+
+    /// Verb IDs that have at least one irregular span in the requested categories.
+    /// Returns an empty set when `categories` is empty — caller decides whether
+    /// that means "no constraint" or "no matches".
+    public static func verbIDs(
+        matchingIrregularCategories categories: Set<IrregularSpan.SpanCategory>,
+        in spans: [IrregularSlot]
+    ) -> Set<Int> {
+        guard !categories.isEmpty else { return [] }
+        var ids = Set<Int>()
+        for slot in spans where categories.contains(slot.category) {
+            ids.insert(slot.verbId)
+        }
+        return ids
+    }
+
+    /// Practice pool: verbs at the selected levels, intersected with irregular
+    /// categories when that filter is active.
+    ///
+    /// Semantics (Issue #26):
+    /// - `selectedLevels` empty → empty pool (literal).
+    /// - `irregularCategories` empty → no irregular constraint (all verbs at level).
+    public static func allowedVerbIDs(
+        verbs: [VerbSlot],
+        spans: [IrregularSlot],
+        selectedLevels: Set<String>,
+        irregularCategories: Set<IrregularSpan.SpanCategory>
+    ) -> Set<Int> {
+        let levelIDs = verbIDs(matchingLevels: selectedLevels, in: verbs)
+        guard !irregularCategories.isEmpty else { return levelIDs }
+        let irregularIDs = verbIDs(matchingIrregularCategories: irregularCategories, in: spans)
+        return levelIDs.intersection(irregularIDs)
+    }
+}

Conjuga/SharedModels/Sources/SharedModels/ReflexiveVerb.swift

@@ -0,0 +1,23 @@
+import Foundation
+
+/// A single entry from the curated "100 most common reflexive verbs" list
+/// (Gitea issue #28). Sourced from spanishwithdaniel.com.
+///
+/// `baseInfinitive` is the stem without the reflexive "-se" suffix, used to
+/// match this entry to the app's Verb records (which store bare infinitives).
+/// `usageHint` captures trailing prepositions or set-phrase completions — e.g.,
+/// "a" for `acercarse a`, "de acuerdo" for `ponerse de acuerdo`. Nil when the
+/// reflexive form has no commonly paired preposition.
+public struct ReflexiveVerb: Codable, Hashable, Sendable {
+    public let infinitive: String
+    public let baseInfinitive: String
+    public let english: String
+    public let usageHint: String?
+
+    public init(infinitive: String, baseInfinitive: String, english: String, usageHint: String? = nil) {
+        self.infinitive = infinitive
+        self.baseInfinitive = baseInfinitive
+        self.english = english
+        self.usageHint = usageHint
+    }
+}

Conjuga/SharedModels/Sources/SharedModels/TextbookChapter.swift

@@ -0,0 +1,86 @@
+import Foundation
+import SwiftData
+
+/// One chapter of the textbook. Ordered content blocks are stored as JSON in `bodyJSON`
+/// (encoded [TextbookBlock]) since SwiftData @Model doesn't support heterogeneous arrays.
+@Model
+public final class TextbookChapter {
+    @Attribute(.unique) public var id: String = ""
+    public var number: Int = 0
+    public var title: String = ""
+    public var part: Int = 0                    // 0 = no part assignment
+    public var courseName: String = ""
+    public var bodyJSON: Data = Data()
+    public var exerciseCount: Int = 0
+    public var vocabTableCount: Int = 0
+
+    public init(
+        id: String,
+        number: Int,
+        title: String,
+        part: Int,
+        courseName: String,
+        bodyJSON: Data,
+        exerciseCount: Int,
+        vocabTableCount: Int
+    ) {
+        self.id = id
+        self.number = number
+        self.title = title
+        self.part = part
+        self.courseName = courseName
+        self.bodyJSON = bodyJSON
+        self.exerciseCount = exerciseCount
+        self.vocabTableCount = vocabTableCount
+    }
+
+    public func blocks() -> [TextbookBlock] {
+        (try? JSONDecoder().decode([TextbookBlock].self, from: bodyJSON)) ?? []
+    }
+}
+
+/// One content block within a chapter. Polymorphic via `kind`.
+public struct TextbookBlock: Codable, Identifiable, Sendable {
+    public enum Kind: String, Codable, Sendable {
+        case heading
+        case paragraph
+        case keyVocabHeader = "key_vocab_header"
+        case vocabTable = "vocab_table"
+        case exercise
+    }
+
+    public var id: String { "\(kind.rawValue):\(index)" }
+    public var index: Int
+    public var kind: Kind
+
+    // heading
+    public var level: Int?
+    // heading / paragraph
+    public var text: String?
+
+    // vocab_table
+    public var sourceImage: String?
+    public var ocrLines: [String]?
+    public var ocrConfidence: Double?
+    public var cards: [TextbookVocabPair]?
+
+    // exercise
+    public var exerciseId: String?
+    public var instruction: String?
+    public var extra: [String]?
+    public var prompts: [String]?
+    public var answerItems: [TextbookAnswerItem]?
+    public var freeform: Bool?
+}
+
+public struct TextbookVocabPair: Codable, Sendable {
+    public var front: String
+    public var back: String
+}
+
+public struct TextbookAnswerItem: Codable, Sendable {
+    public var label: String?      // A/B/C subpart label or nil
+    public var number: Int
+    public var answer: String
+    public var alternates: [String]
+}

Conjuga/SharedModels/Sources/SharedModels/TextbookExerciseAttempt.swift

@@ -0,0 +1,83 @@
+import Foundation
+import SwiftData
+
+/// Per-prompt grading state recorded after the user submits an exercise.
+public enum TextbookGrade: Int, Codable, Sendable {
+    case wrong = 0
+    case close = 1
+    case correct = 2
+}
+
+/// User's attempt for one exercise. Stored in the cloud container so progress
+/// syncs across devices.
+@Model
+public final class TextbookExerciseAttempt {
+    /// Deterministic id: "<courseName>|<exerciseId>". CloudKit-synced models can't
+    /// use @Attribute(.unique); code that writes attempts must fetch-or-create.
+    public var id: String = ""
+    public var courseName: String = ""
+    public var chapterNumber: Int = 0
+    public var exerciseId: String = ""
+
+    /// JSON-encoded per-prompt state array.
+    /// Each entry: { "number": Int, "userText": String, "grade": Int }
+    public var stateJSON: Data = Data()
+
+    public var lastAttemptAt: Date = Date()
+    public var correctCount: Int = 0
+    public var closeCount: Int = 0
+    public var wrongCount: Int = 0
+    public var totalCount: Int = 0
+
+    public init(
+        id: String,
+        courseName: String,
+        chapterNumber: Int,
+        exerciseId: String,
+        stateJSON: Data = Data(),
+        lastAttemptAt: Date = Date(),
+        correctCount: Int = 0,
+        closeCount: Int = 0,
+        wrongCount: Int = 0,
+        totalCount: Int = 0
+    ) {
+        self.id = id
+        self.courseName = courseName
+        self.chapterNumber = chapterNumber
+        self.exerciseId = exerciseId
+        self.stateJSON = stateJSON
+        self.lastAttemptAt = lastAttemptAt
+        self.correctCount = correctCount
+        self.closeCount = closeCount
+        self.wrongCount = wrongCount
+        self.totalCount = totalCount
+    }
+
+    public func promptStates() -> [TextbookPromptState] {
+        (try? JSONDecoder().decode([TextbookPromptState].self, from: stateJSON)) ?? []
+    }
+
+    public func setPromptStates(_ states: [TextbookPromptState]) {
+        stateJSON = (try? JSONEncoder().encode(states)) ?? Data()
+        correctCount = states.filter { $0.grade == .correct }.count
+        closeCount = states.filter { $0.grade == .close }.count
+        wrongCount = states.filter { $0.grade == .wrong }.count
+        totalCount = states.count
+    }
+
+    public static func attemptId(courseName: String, exerciseId: String) -> String {
+        "\(courseName)|\(exerciseId)"
+    }
+}
+
+public struct TextbookPromptState: Codable, Sendable {
+    public var number: Int
+    public var userText: String
+    public var grade: TextbookGrade
+
+    public init(number: Int, userText: String, grade: TextbookGrade) {
+        self.number = number
+        self.userText = userText
+        self.grade = grade
+    }
+}

Conjuga/SharedModels/Sources/SharedModels/VerbExample.swift

@@ -0,0 +1,17 @@
+import Foundation
+
+/// A single Spanish / English example sentence pair for a verb at a specific tense.
+/// Used by the Verb detail view (Issue #27). Generated at runtime via Foundation
+/// Models and cached to disk; shape is intentionally simple Codable for easy
+/// JSON persistence and cross-module sharing.
+public struct VerbExample: Codable, Hashable, Sendable {
+    public let tenseId: String
+    public let spanish: String
+    public let english: String
+
+    public init(tenseId: String, spanish: String, english: String) {
+        self.tenseId = tenseId
+        self.spanish = spanish
+        self.english = english
+    }
+}

Conjuga/SharedModels/Sources/SharedModels/VerbLevel+Highest.swift

@@ -0,0 +1,10 @@
+import Foundation
+
+public extension VerbLevel {
+    /// The highest-ranked `VerbLevel` in `set` per `allCases` ordering.
+    /// Used when a single representative level is required (word-of-day
+    /// widget, AI chat/story scenario generation).
+    static func highest(in set: Set<VerbLevel>) -> VerbLevel? {
+        allCases.last { set.contains($0) }
+    }
+}

Conjuga/SharedModels/Tests/SharedModelsTests/AnswerGraderTests.swift

@@ -0,0 +1,80 @@
+import Testing
+@testable import SharedModels
+
+@Suite("AnswerGrader")
+struct AnswerGraderTests {
+
+    @Test("exact match is correct")
+    func exact() {
+        #expect(AnswerGrader.grade(userText: "tengo", canonical: "tengo") == .correct)
+        #expect(AnswerGrader.grade(userText: "Tengo", canonical: "tengo") == .correct)
+        #expect(AnswerGrader.grade(userText: "  tengo  ", canonical: "tengo") == .correct)
+    }
+
+    @Test("missing accent is close")
+    func missingAccent() {
+        #expect(AnswerGrader.grade(userText: "esta", canonical: "está") == .close)
+        #expect(AnswerGrader.grade(userText: "nino", canonical: "niño") == .close)
+        #expect(AnswerGrader.grade(userText: "asi", canonical: "así") == .close)
+    }
+
+    @Test("single-char typo is close")
+    func singleCharTypo() {
+        // deletion
+        #expect(AnswerGrader.grade(userText: "tngo", canonical: "tengo") == .close)
+        // insertion
+        #expect(AnswerGrader.grade(userText: "tengoo", canonical: "tengo") == .close)
+        // substitution
+        #expect(AnswerGrader.grade(userText: "tengu", canonical: "tengo") == .close)
+    }
+
+    @Test("two-char typo is wrong")
+    func twoCharTypo() {
+        #expect(AnswerGrader.grade(userText: "tngu", canonical: "tengo") == .wrong)
+    }
+
+    @Test("empty is wrong")
+    func empty() {
+        #expect(AnswerGrader.grade(userText: "", canonical: "tengo") == .wrong)
+        #expect(AnswerGrader.grade(userText: "   ", canonical: "tengo") == .wrong)
+    }
+
+    @Test("alternates accepted")
+    func alternates() {
+        #expect(AnswerGrader.grade(userText: "flaca", canonical: "delgada", alternates: ["flaca"]) == .correct)
+        #expect(AnswerGrader.grade(userText: "flacca", canonical: "delgada", alternates: ["flaca"]) == .close)
+    }
+
+    @Test("punctuation stripped")
+    func punctuation() {
+        #expect(AnswerGrader.grade(userText: "el libro.", canonical: "el libro") == .correct)
+        #expect(AnswerGrader.grade(userText: "¿dónde?", canonical: "dónde") == .correct)
+    }
+
+    @Test("very different text is wrong")
+    func wrong() {
+        #expect(AnswerGrader.grade(userText: "hola", canonical: "tengo") == .wrong)
+        #expect(AnswerGrader.grade(userText: "casa", canonical: "perro") == .wrong)
+    }
+
+    @Test("normalize produces expected output")
+    func normalize() {
+        #expect(AnswerGrader.normalize("  Hola  ") == "hola")
+        #expect(AnswerGrader.normalize("ABC!") == "abc")
+    }
+
+    @Test("stripAccents handles common Spanish diacritics")
+    func stripAccents() {
+        #expect(AnswerGrader.stripAccents("niño") == "nino")
+        #expect(AnswerGrader.stripAccents("está") == "esta")
+        #expect(AnswerGrader.stripAccents("güero") == "guero")
+    }
+
+    @Test("levenshtein computes edit distance")
+    func levenshtein() {
+        #expect(AnswerGrader.levenshtein("kitten", "sitting") == 3)
+        #expect(AnswerGrader.levenshtein("flaw", "lawn") == 2)
+        #expect(AnswerGrader.levenshtein("abc", "abc") == 0)
+        #expect(AnswerGrader.levenshtein("", "abc") == 3)
+    }
+}

Conjuga/SharedModels/Tests/SharedModelsTests/PracticeFilterTests.swift

@@ -0,0 +1,212 @@
+import Testing
+@testable import SharedModels
+
+// Practice pool = selected levels ∩ selected tenses ∩ selected irregular types
+// (Issue #26). This suite covers the level + irregular intersection; tense
+// filtering is a separate concern handled at the review-card layer.
+
+@Suite("PracticeFilter — level selection")
+struct PracticeFilterLevelTests {
+
+    @Test("single level expands to that group's data-levels")
+    func singleLevelExpansion() {
+        let expected = VerbLevelGroup.dataLevels(for: "elementary")
+        #expect(PracticeFilter.dataLevels(forSelectedLevels: ["elementary"]) == expected)
+    }
+
+    @Test("multi-level union merges every group's data-levels")
+    func multiLevelUnion() {
+        let union = PracticeFilter.dataLevels(forSelectedLevels: ["elementary", "intermediate"])
+        #expect(union.isSuperset(of: VerbLevelGroup.dataLevels(for: "elementary")))
+        #expect(union.isSuperset(of: VerbLevelGroup.dataLevels(for: "intermediate")))
+    }
+
+    @Test("empty selected-levels produces empty data-levels")
+    func emptyLevelsProducesEmpty() {
+        #expect(PracticeFilter.dataLevels(forSelectedLevels: []).isEmpty)
+    }
+
+    @Test("unknown level passes through to its raw value")
+    func unknownLevelPassthrough() {
+        // Preserves VerbLevelGroup's fallback contract.
+        #expect(PracticeFilter.dataLevels(forSelectedLevels: ["custom"]) == ["custom"])
+    }
+}
+
+@Suite("PracticeFilter — verb IDs by level")
+struct PracticeFilterVerbLevelTests {
+
+    // Fixtures: four verbs spanning basic + elementary subgroups + intermediate.
+    private let verbs: [PracticeFilter.VerbSlot] = [
+        .init(id: 1, level: "basic"),
+        .init(id: 2, level: "elementary"),
+        .init(id: 3, level: "elementary_2"),
+        .init(id: 4, level: "intermediate_1"),
+    ]
+
+    @Test("elementary selection matches elementary base and subgroup levels")
+    func elementarySubgroupMatch() {
+        let ids = PracticeFilter.verbIDs(matchingLevels: ["elementary"], in: verbs)
+        #expect(ids == [2, 3])
+    }
+
+    @Test("multi-level selection unions matching verb IDs")
+    func multiLevelUnionIds() {
+        let ids = PracticeFilter.verbIDs(matchingLevels: ["basic", "intermediate"], in: verbs)
+        #expect(ids == [1, 4])
+    }
+
+    @Test("empty level selection returns no verbs (literal semantics)")
+    func emptySelectionReturnsEmpty() {
+        let ids = PracticeFilter.verbIDs(matchingLevels: [], in: verbs)
+        #expect(ids.isEmpty)
+    }
+}
+
+@Suite("PracticeFilter — verb IDs by irregular category")
+struct PracticeFilterIrregularTests {
+
+    // Verb 10: regular (no spans).
+    // Verb 20: one spelling-change span.
+    // Verb 30: stem-change + unique-irregular spans.
+    private let spans: [PracticeFilter.IrregularSlot] = [
+        .init(verbId: 20, category: .spelling),
+        .init(verbId: 30, category: .stemChange),
+        .init(verbId: 30, category: .uniqueIrregular),
+    ]
+
+    @Test("empty category set returns empty — caller decides the semantics")
+    func emptyCategoriesReturnsEmpty() {
+        let ids = PracticeFilter.verbIDs(matchingIrregularCategories: [], in: spans)
+        #expect(ids.isEmpty)
+    }
+
+    @Test("single category picks matching verbs only")
+    func singleCategoryMatch() {
+        let ids = PracticeFilter.verbIDs(matchingIrregularCategories: [.spelling], in: spans)
+        #expect(ids == [20])
+    }
+
+    @Test("multiple categories union their matches")
+    func multipleCategoriesUnion() {
+        let ids = PracticeFilter.verbIDs(
+            matchingIrregularCategories: [.spelling, .stemChange],
+            in: spans
+        )
+        #expect(ids == [20, 30])
+    }
+
+    @Test("a verb with multiple matching spans is returned once")
+    func verbWithMultipleSpansDeduped() {
+        let ids = PracticeFilter.verbIDs(
+            matchingIrregularCategories: [.stemChange, .uniqueIrregular],
+            in: spans
+        )
+        #expect(ids == [30])
+    }
+}
+
+@Suite("PracticeFilter — allowedVerbIDs (levels ∩ irregulars)")
+struct PracticeFilterIntersectionTests {
+
+    // Realistic fixture:
+    // #1 — basic, regular
+    // #2 — basic, spelling-change
+    // #3 — elementary, spelling-change
+    // #4 — elementary, stem-change
+    // #5 — intermediate, unique-irregular
+    private let verbs: [PracticeFilter.VerbSlot] = [
+        .init(id: 1, level: "basic"),
+        .init(id: 2, level: "basic"),
+        .init(id: 3, level: "elementary"),
+        .init(id: 4, level: "elementary_1"),
+        .init(id: 5, level: "intermediate"),
+    ]
+    private let spans: [PracticeFilter.IrregularSlot] = [
+        .init(verbId: 2, category: .spelling),
+        .init(verbId: 3, category: .spelling),
+        .init(verbId: 4, category: .stemChange),
+        .init(verbId: 5, category: .uniqueIrregular),
+    ]
+
+    @Test("no irregular filter keeps every verb at the selected level")
+    func noIrregularConstraint() {
+        let ids = PracticeFilter.allowedVerbIDs(
+            verbs: verbs,
+            spans: spans,
+            selectedLevels: ["basic"],
+            irregularCategories: []
+        )
+        #expect(ids == [1, 2])
+    }
+
+    @Test("Issue #26 worked example: beginner + spelling-change → only #2")
+    func issueWorkedExample() {
+        let ids = PracticeFilter.allowedVerbIDs(
+            verbs: verbs,
+            spans: spans,
+            selectedLevels: ["basic"],
+            irregularCategories: [.spelling]
+        )
+        #expect(ids == [2])
+    }
+
+    @Test("filter is an intersection, not a union: level-mismatched spans are excluded")
+    func intersectionExcludesOtherLevels() {
+        // Elementary has a spelling-change verb (#3). Selecting basic + spelling
+        // must NOT leak #3 through the irregular filter alone.
+        let ids = PracticeFilter.allowedVerbIDs(
+            verbs: verbs,
+            spans: spans,
+            selectedLevels: ["basic"],
+            irregularCategories: [.spelling]
+        )
+        #expect(!ids.contains(3))
+    }
+
+    @Test("empty level selection produces empty pool regardless of irregular filter")
+    func emptyLevelsLocksOutPractice() {
+        let ids = PracticeFilter.allowedVerbIDs(
+            verbs: verbs,
+            spans: spans,
+            selectedLevels: [],
+            irregularCategories: [.spelling]
+        )
+        #expect(ids.isEmpty)
+    }
+
+    @Test("multi-level + multi-category pulls every matching pair")
+    func multiLevelMultiCategory() {
+        let ids = PracticeFilter.allowedVerbIDs(
+            verbs: verbs,
+            spans: spans,
+            selectedLevels: ["elementary", "intermediate"],
+            irregularCategories: [.stemChange, .uniqueIrregular]
+        )
+        #expect(ids == [4, 5])
+    }
+}
+
+@Suite("VerbLevel.highest")
+struct VerbLevelHighestTests {
+
+    @Test("returns the highest-ranked level in the set")
+    func highestOfMany() {
+        #expect(VerbLevel.highest(in: [.basic, .intermediate, .elementary]) == .intermediate)
+    }
+
+    @Test("returns the sole element when set is a singleton")
+    func singleton() {
+        #expect(VerbLevel.highest(in: [.advanced]) == .advanced)
+    }
+
+    @Test("returns nil for the empty set")
+    func empty() {
+        #expect(VerbLevel.highest(in: []) == nil)
+    }
+
+    @Test("ranks expert above advanced above intermediate above elementary above basic")
+    func fullRanking() {
+        #expect(VerbLevel.highest(in: [.basic, .elementary, .intermediate, .advanced, .expert]) == .expert)
+    }
+}

Conjuga/project.yml

@@ -47,6 +47,8 @@ targets:
         buildPhase: resources
       - path: Conjuga/course_data.json
         buildPhase: resources
+      - path: Conjuga/reflexive_verbs.json
+        buildPhase: resources
     info:
       path: Conjuga/Info.plist
       properties:

README.md

@@ -1,32 +1,57 @@
 # Conjuga
 
-A Spanish verb conjugation learning app for iOS. Practice all 20 tenses with spaced repetition, handwriting recognition, and vocabulary courses.
+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
 
-- **Conjugation Practice** — Flashcards, typing, multiple choice, handwriting, sentence building, and full table modes
+### Verb practice
-- **Spaced Repetition** — SM-2 algorithm tracks what you know and surfaces what you don't
+- **Six conjugation modes** — flashcards, typing, multiple choice, handwriting (Apple Pencil / finger), sentence builder, and full-table (conjugate all persons at once)
-- **1,750 Verbs** — From basic to expert, with irregular form highlighting
+- **Focus modes** — weak verbs (SM-2 spaced repetition), irregularity drills (spelling / stem / unique), common tenses
-- **20 Tenses** — Every indicative, subjunctive, conditional, and imperative tense
+- **1,750 verbs** across 5 levels (Basic → Expert) with 209 K pre-conjugated forms and 23 K irregular-span annotations
-- **Grammar Guide** — 20 tense guides with conjugation tables + 20 grammar topic notes (Ser vs Estar, Por vs Para, etc.)
+- **20 tenses** — every indicative, subjunctive, conditional, and imperative tense, each with character-level irregular highlighting
-- **Vocabulary Courses** — Weekly decks with example sentences
+- **Irregularity filter** — search the verb list by Any Irregular / Spelling Change / Stem Change / Unique Irregular, combinable with level filter
-- **Progress Tracking** — Streaks, daily goals, accuracy stats, and achievement badges
+- **Text-to-speech** on any form
-- **CloudKit Sync** — Review progress syncs across devices
+
-- **Widgets** — Daily progress, word of the day, and weekly stats
+### Content & study
-- **Text-to-Speech** — Hear any verb pronounced in Spanish
+- **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 + 20+ grammar topic notes (ser/estar, por/para, preterite/imperfect, etc.), each with 100+ practice exercises
+- **Grammar exercises** — interactive quizzes for 5 core topics (ser/estar, por/para, preterite/imperfect, subjunctive, personal *a*)
+
+### 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
+
+### 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** — Reference data (verbs, forms, guides) seeded from JSON on first launch
+  - **Local store** (App Group `group.com.conjuga.app`) — reference data: verbs, forms, irregular spans, tense guides, course decks, vocab cards, textbook chapters. Seeded from bundled JSON on first launch. Self-healing re-seeds trigger on version bumps *or* if rows are missing on disk.
-  - **Cloud store** — User data (review cards, progress, streaks) synced via CloudKit
+  - **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** package shared between the app and widget extension
+- **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).
+- **Vision** framework for OCR of textbook pages and vocabulary images.
+- **Speech** framework for recognition and pronunciation scoring.
+- **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.
 
 ## Requirements
 
-- iOS 18+
+- iOS 18+ (iOS 26 for Foundation Models features)
 - Xcode 16+
 
 ## Building
 
-Open `Conjuga/Conjuga.xcodeproj` in Xcode and run on a simulator or device. Data seeds automatically on first launch.
+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.

app_features.md

@@ -1,6 +1,85 @@
 # Spanish Conjugation Apps — Feature Comparison
 
-Side-by-side feature analysis of **ConjuGato** and **Conjuu ES**, based on App Store screenshots and extracted app data.
+Side-by-side feature analysis of **Conjuga** (this project), **ConjuGato**, and **Conjuu ES**, based on App Store screenshots, extracted app data, and this repository's source.
+
+---
+
+## Conjuga (this project)
+
+**Platform:** iOS 18+ (iOS 26 required for Foundation Models features)
+**Monetization:** —
+**Tech stack:** SwiftUI + SwiftData (dual local / CloudKit stores), SharedModels Swift Package, Foundation Models, Vision, Speech, WidgetKit
+
+### Practice Modes
+
+- **Six core conjugation modes** — flashcards, typing, multiple choice, handwriting (Apple Pencil + finger), sentence builder, full table (all persons at once)
+- **Focus modes** — weak verbs (SM-2 SRS), irregularity drills (spelling / stem / unique, selectable), common tenses
+- **Quick answer review** with per-form irregular-span highlighting
+- **Vocab SRS Review** — spaced repetition over course vocabulary with Again / Hard / Good / Easy rating
+- **Cloze practice** — fill-in-the-blank with distractor generation from vocab pool
+- **Listening practice** — listen-and-type + pronunciation scoring via Speech framework, word-by-word match
+- **Lyrics practice** — search Spanish songs, translate line by line
+- **Conversational practice** — on-device AI chat partner (Apple Foundation Models) with 10 scenarios, tappable words that open dictionary or on-demand AI lookup
+- **AI short stories** — generated stories with tappable words + comprehension quiz
+
+### Verb Reference
+
+- **1,750 verbs**, 5 levels (Basic / Elementary / Intermediate / Advanced / Expert), 209 K pre-conjugated forms
+- **20 tenses** with character-level irregular-span highlighting (spelling / stem / unique)
+- **Irregularity filter** — All / Any Irregular / Spelling Change / Stem Change / Unique Irregular — combinable with level
+- **Per-verb detail** with per-form English translations and tense-grouped table
+- **Text-to-speech** on any conjugated form
+
+### Grammar & Content
+
+- **20 tense guides** with usage rules and examples
+- **20+ grammar topic notes** (ser/estar, por/para, preterite/imperfect, subjunctive, personal *a*, suffixes, irregular yo forms, stem-changers, etc.) each with 100+ practice exercises
+- **Grammar exercises** — interactive quizzes for 5 core topics
+- **Course decks** — weekly vocabulary with example sentences, week tests, cumulative checkpoint exams
+- **Stem-change toggle** on Week 4 decks (E-IE, E-I, O-UE, U-UE) with inline present-tense conjugations
+- **Textbook reader** — 30 chapters of *Complete Spanish Step-by-Step* with 251 interactive exercises (keyboard + Apple Pencil), 931 OCR'd vocab tables rendered as paired Spanish→English grids (~3 100 cards)
+- **Textbook extraction pipeline** — XHTML + answer-key parsers, macOS Vision image OCR, PDF page OCR, bounding-box vocab pair extractor, NSSpellChecker validator, language-aware auto-fixer
+
+### Offline Dictionary
+
+- **Reverse index of 175 K verb forms + 200 common words**, cached to disk
+- **Tappable-word lookup** in chat bubbles and stories; falls back to Foundation Models `@Generable ChatWordInfo` when a word isn't in the dictionary
+
+### Progress & Sync
+
+- **SM-2 spaced repetition** for verb review
+- **Streaks, daily goals, accuracy stats, achievement badges**
+- **Study-time tracking** per day (foreground time)
+- **CloudKit private-database sync** — review cards, user progress, test results, daily logs, saved songs, stories, conversations, textbook exercise attempts
+- **Background app refresh** for widget data
+
+### Widgets
+
+- **Combined dashboard** (word of day + stats)
+- **Word-of-day** standalone
+- Both refresh daily and on app backgrounding
+
+### Settings & Filters
+
+- **Selectable verb level** and **enabled tenses**
+- **Include vosotros** toggle
+- **Auto-fill verb stem** toggle for Full Table practice
+- **Feature reference** page in Settings documenting every feature and which settings affect it
+
+### Data (in repo)
+
+| Asset | Count |
+|-------|-------|
+| Verbs | 1,750 |
+| Verb forms (pre-conjugated) | 209,014 |
+| Irregular span annotations | 23,795 |
+| Tenses | 20 |
+| Tense guides | 20 |
+| Grammar notes | 20+ (each with 100+ exercises) |
+| Textbook chapters | 30 |
+| Textbook exercises | 251 |
+| Textbook vocab pairs | ~3,118 |
+| Offline dictionary forms | 175,425 |
 
 ---
 
@@ -184,30 +263,44 @@ Side-by-side feature analysis of **ConjuGato** and **Conjuu ES**, based on App S
 
 ## Feature Comparison
 
-| Feature | ConjuGato | Conjuu ES |
+| Feature | Conjuga | ConjuGato | Conjuu ES |
-|---------|-----------|-----------|
+|---------|---------|-----------|-----------|
-| **Verb count** | 1,750 | 621 |
+| **Verb count** | 1,750 | 1,750 | 621 |
-| **Tenses** | 27 | 20 |
+| **Tenses** | 20 | 27 | 20 |
-| **Practice style** | Flashcard (tap to reveal) | Typing (fill in forms) |
+| **Practice style** | Flashcards, typing, MC, handwriting, sentence builder, full table | Flashcard (tap to reveal) | Typing (fill in forms) |
-| **Input method** | Self-rate (no typing) | Type conjugation |
+| **Input method** | All four (type / tap / write / speak) | Self-rate (no typing) | Type conjugation |
-| **Daily goals** | Numeric counter (25/100) | 4 presets (Casual→Committed) |
+| **Daily goals** | Numeric counter + streaks + study-time | Numeric counter (25/100) | 4 presets (Casual→Committed) |
-| **Spaced repetition** | Yes (star + heart ratings) | Session-based review |
+| **Spaced repetition** | SM-2 (verb forms + vocab cards) | Yes (star + heart ratings) | Session-based review |
-| **Rhyming mnemonics** | Yes (dedicated tab) | Not visible |
+| **Grammar breakdown** | Character-level irregular span highlighting | Color-coded span highlighting | Ending table + usage rules |
-| **Grammar breakdown** | Color-coded span highlighting | Ending table + usage rules |
+| **Irregularity detail** | 3-level (spelling/stem/unique) + per-verb badges + filter | 3-level with character spans | 4-label (ordinary/irregular/regular/orto) |
-| **Irregularity detail** | 3-level (spelling/stem/unique) with character spans | 4-label (ordinary/irregular/regular/orto) |
+| **Tense explanations** | 20 guides + 20 grammar notes with 100+ exercises each | In-app guides with mnemonics | Modal popups with usage categories |
-| **Tense explanations** | In-app guides with mnemonics | Modal popups with usage categories |
+| **Example sentences** | Per-tense + per-card + per-story + in textbook | Per-tense with audio | Per-tense in guide popups |
-| **Example sentences** | Per-tense with audio | Per-tense in guide popups |
+| **Audio** | TTS on any form + pronunciation scoring | Tap any conjugation + auto-pronounce | Not prominent |
-| **Audio** | Tap any conjugation + auto-pronounce | Not prominent |
+| **Difficulty levels** | 5 levels + irregularity filter (combinable) | Filter by irregularity type | 8 graded word lists |
-| **Difficulty levels** | Filter by irregularity type | 8 graded word lists |
+| **Custom lists** | — | No | Yes (My Lists + Pinboard) |
-| **Custom lists** | No | Yes (My Lists + Pinboard) |
+| **Verb filters** | Level + irregularity category + tense selection + search | Ending type + irregularity + tense | Tense checkboxes + level |
-| **Verb filters** | Ending type + irregularity + tense | Tense checkboxes + level |
+| **AI features** | On-device chat, stories, word lookups (Foundation Models) | — | — |
-| **Translations** | 8 languages | 4 languages |
+| **Textbook** | 30 chapters, 251 exercises, ~3,100 vocab pairs (OCR pipeline) | — | — |
-| **Platform** | iOS (Catalyst) | macOS + iOS |
+| **Listening / speech** | Listen-and-type + pronunciation scoring | — | — |
-| **Monetization** | One-time purchase | Free tier + paid |
+| **Offline dictionary** | 175 K forms | — | — |
-| **vosotros toggle** | Yes | Not visible |
+| **Lyrics** | Search + translate | — | — |
+| **CloudKit sync** | Yes (private database) | — | — |
+| **Widgets** | Combined + word-of-day | — | — |
+| **Platform** | iOS 18+ (iOS 26 for AI) | iOS (Catalyst) | macOS + iOS |
+| **Monetization** | — | One-time purchase | Free tier + paid |
+| **vosotros toggle** | Yes | Yes | Not visible |
 
 ## Unique Strengths
 
+### Conjuga excels at:
+- **Breadth of modes** — six conjugation practice styles plus cloze, listening, lyrics, chat, stories, textbook, and vocab review in one app
+- **AI integration** — on-device Foundation Models power the chat partner, story generator, and dictionary fallback lookup with no cloud round-trips
+- **Textbook reader** — full *Complete Spanish Step-by-Step* textbook with 251 interactive exercises and 3,100 OCR-paired vocab cards, built via an in-repo extraction pipeline
+- **Combinable filters** — Level × Irregularity × search on the verb list, with per-row category badges
+- **Multi-input practice** — type, tap, Apple Pencil handwriting, and voice (pronunciation scoring)
+- **Offline dictionary** — 175 K verb-form reverse index makes word lookups instant and network-free
+- **CloudKit sync** — progress, tests, saved content, and conversations travel between devices
+
 ### ConjuGato excels at:
 - **Breadth** — nearly 3× the verbs
 - **Irregularity teaching** — character-level color-coded highlighting showing exactly *why* each form is irregular