ACTA TEE Assistant — native iOS SwiftUI app for CT anesthesiology fellows

Point-of-care TEE case logger. Collapses a ~5-minute logging workflow into a ~40-second capture and tracks progress against ACGME / NBE (Advanced PTEeXAM, Basic PTE) case-minimum requirements.

Hard constraints

• Native iOS only. No web components (no WKWebView, no HTML/JS, no React Native/Flutter), no backend, no cloud sync, no URLSession networking, no third-party packages. Everything is Apple system frameworks.
• De-identified by design: no name / MRN / DOB / hospital is ever stored (age / sex / BSA only).

Repository layout

Package.swift              # SwiftPM: ACTATEEAssistantCore (pure Swift, Linux-testable) + tests
project.yml                # XcodeGen project definition (do NOT hand-edit .xcodeproj)
ACTATEEAssistant/                    # iOS app target
  Models/                  #   SwiftData @Model entities (CaseLog, ValveFinding, QuizCard)
  Engine/                  #   Apple-only engine glue (SpeechController, PDF renderers,
                           #   QuizCardEnqueuer) — pure core engines compile into ACTATEEAssistantCore
  Intents/                 #   App Intents (F1): LogTEE / ShowProgress / QuickFindings
  Shared/                  #   ModelContainerFactory — app-group store shared with widget
  ViewModels/              #   @Observable view models (MVVM)
  Views/                   #   SwiftUI screens + shared components (incl. Quiz/, Insights/)
  Info.plist               #   usage strings (§10) — generated by XcodeGen from project.yml
  ACTATEEAssistant.entitlements      #   app group (F1) — no network, no iCloud
ACTATEEWidget/                 # WidgetKit extension (F1): total/totalMinimum complication
Tests/ACTATEEAssistantCoreTests/     # table-driven unit tests (run with `swift test`)

Core/tests are NOT Apple-only. ACTATEEAssistant/Models + Engine (enums, tracks, requirements engine, auto-categorizer, dictation mapper, CSV writer) imports only Foundation, so the engine logic is unit-testable on Linux with swift test — no Xcode required:

$ swift test

On this Silverblue host (no system toolchain), tests run via rootless podman with the swift.org tarball mounted in:

$ podman build -t teelog-swift -f ~/swift/Dockerfile ~/swift   # once
$ podman run --rm -v ~/swift:/swift:ro,Z -v "$PWD":/work:Z -w /work \
      teelog-swift /swift/usr/bin/swift test

Result: 114 tests green (AutoCategorizer ×18, DictationMapper ×32, RequirementsEngine ×16, SpacedRepetition ×11, QuizCardGenerator ×17, InsightsEngine ×20).

Building the app (on a Mac)

1. Install XcodeGen (brew install xcodegen).
2. xcodegen generate in this directory → produces ACTATEEAssistant.xcodeproj.
3. Open ACTATEEAssistant.xcodeproj, select the ACTATEEAssistant scheme, run on an iPhone simulator/device (iOS 17+).
4. Unit tests: xcodebuild test -scheme ACTATEEAssistant -destination 'platform=iOS Simulator,name=iPhone 15' or just swift test for the core engine.

The .xcodeproj is generated and git-ignored — project.yml is the source of truth.

Phase status (v1.0 complete)

• P1 Skeleton — done (models, enums, tab shell, project definition)
• P2 Capture — done (Quick Log 3-step flow, Findings, auto-categorize, Dashboard + Library read views)
• P3 Requirements — done (engine, Progress screen, gap alerts)
• P4 Dictation — done (engine + mapper + SpeechController; DictationSheet with live chip fill while listening, confirmation diff, field-level mics on Procedure / Indication / Views / LV EF / RV / Valve lesions / Notes, speech-not-authorized + on-device-unavailable states with manual fallback)
• P5 Export — done (CSV writer; ACGME / NBE / PDF content builders; multi-page UIGraphicsPDFRenderer with wrapping + page breaks + footers; academic-year range; native preview; share sheet + Save PDF to Files)
• P6 Hardening — done (Face ID / passcode lock with settings toggle and background re-lock; VoiceOver labels incl. severity steppers and progress bars; Dynamic Type safe; ≥44 pt targets; Reduce Motion respected; system colors only; empty states on Dashboard, Library, Progress, Export, Findings)

v1.1 status

• F3 Spaced-repetition quiz — done (SM-2 scheduler + question generator in ACTATEEAssistantCore; QuizCard model; Review · N cards due entry on the Library header; ReviewSessionView with Again/Good/Easy; cards auto-enqueued on case save, deduped per case)
• F1 Siri Shortcuts / hands-free capture — done (LogTEECaseIntent, ShowProgressIntent, QuickFindingsIntent + AppShortcuts provider; WidgetKit extension showing total/totalMinimum with actatee://quicklog tap-through to Quick Log; reload on every case save; shared app-group store + preferences)
• F2 Personal pattern insights — done (InsightsEngine aggregations + InsightsEngineTests; Insights screen from the Progress header with Swift Charts cards, one-line takeaways, tap-through to the filtered Library)

See IMPLEMENTATION.md (§12–§15) for phase details and open product questions (track minima placeholders marked in Tracks.swift).