ActionBrief

Turn messy notes into a clear action plan — entirely on your iPhone.

ActionBrief is a privacy-first iOS app that converts meeting notes, class notes, and personal brain dumps into a structured checklist — complete with owners, due dates, priorities, and a one-sentence summary. It uses on-device Apple Foundation Models when available, and falls back to local NaturalLanguage parsing when they are not.

This is not a chatbot. It solves one practical problem: extracting actionable tasks from unstructured text.

Demo

To embed the video: Go to github.com/devendrabhumca12/ActionBrief, click the pencil ✏️ to edit this README, drag screenshots/actionBrief_compressed.mp4 into the editor — GitHub will upload it and give you a URL to paste here.

https://github.com/devendrabhumca12/ActionBrief/raw/master/screenshots/actionBrief_compressed.mp4

Paste chaotic notes → get a clean action plan → everything stays on device.

Screenshots

Welcome	Input Methods	Extract Actions	Editor

Onboarding with value props	Four input modes explained	AI extraction capabilities	Live editor, On-Device AI Ready

The Problem It Solves

People leave meetings and lectures with rough notes but no clear next steps. Important tasks, dates, and follow-ups get buried in free-form text.

Before — raw notes:

Sprint planning - mobile launch
Raj to check budget before Friday.
I should send the final deck after he replies.
Need to book user testing for next Tuesday.

After — structured brief:

Task	Owner	Due	Priority
Check budget	Raj	Friday	🔴 High
Send final deck	Me	After Raj replies	🔴 High
Book user testing	—	Next Tuesday	🟡 Medium

Summary: Prepare launch materials, validate budget, schedule user testing.

All processing stays on device. No account. No cloud AI. No analytics.

Features

Paste or type notes and extract action items in one tap
Record a spoken transcript with on-device speech recognition
Upload meeting transcript files (.txt, .md, plain text)
Import from photos — OCR text from notes, screens, or whiteboards via Vision
On-device AI via Apple Foundation Models with guided structured output (@Generable)
Automatic fallback using NaturalLanguage + NSDataDetector on unsupported devices
Editable checklist with owners, due dates, and priority labels
Local JSON persistence — saved briefs survive app restarts
Native share sheet to export extracted actions
Clear AI availability states — users always know what mode they're in
Accessibility-first — VoiceOver labels and Dynamic Type throughout
Zero permissions in v1.0 — no tracking, no network, no data collection

Tech Stack

Layer	Technology
UI Framework	SwiftUI
Architecture	MVVM
Concurrency	Swift async/await
On-device AI	Apple Foundation Models (`@Generable`, `LanguageModelSession`)
Fallback parsing	`NaturalLanguage`, `NSDataDetector`
Speech input	`Speech` framework (on-device recognition)
Photo OCR	`Vision` (`VNRecognizeTextRequest`)
File import	`fileImporter` (SwiftUI)
Storage	JSON in Application Support
Minimum iOS	18.0
Full AI support	iOS 26.0+ with Apple Intelligence
Xcode	26+

Architecture

Views  →  ViewModels  →  Services  →  Models

flowchart TD
    UserInput[User enters notes] --> EditorVM[BriefEditorViewModel]
    EditorVM --> AvailabilityCheck[AIAvailabilityService]
    AvailabilityCheck -->|Available| AIService[FoundationModelExtractionService]
    AvailabilityCheck -->|Fallback| FallbackService[FallbackExtractionService]
    AIService --> Result[ExtractionResult]
    FallbackService --> Result
    Result --> ReviewUI[ActionReviewView]
    ReviewUI --> Store[BriefStore]
    Store --> SavedList[SavedBriefsView]

Key Components

File	Role
`BriefEditorViewModel`	Note input, extraction state machine, error handling
`ActionExtractionService`	Protocol that abstracts AI and fallback paths
`FoundationModelExtractionService`	Guided on-device structured generation (`iOS 26+`)
`FallbackExtractionService`	Sentence-level parsing + date detection (`iOS 18+`)
`AIAvailabilityService`	Maps platform availability to user-facing states
`BriefStore`	JSON persistence in Application Support
`ExtractionServiceFactory`	Chooses the right service at runtime

AI Workflow

Primary Path (iOS 26+, Apple Intelligence enabled)

Check SystemLanguageModel.default.availability
Create a LanguageModelSession with structured instructions
Generate typed output using @Generable — no prompt engineering hacks
Map GeneratedBrief → [ActionItem] models with confidence scores

Fallback Path (iOS 18–25 or AI disabled)

Tokenize text into sentences with NLTokenizer
Detect action verbs and date phrases with NSDataDetector
Infer owner and priority via lightweight heuristics
Return lower-confidence results with transparent messaging

The app never uses Private Cloud Compute, third-party AI APIs, or any network connection for note processing.

Requirements

macOS with Xcode 26 Beta or later
iPhone or Simulator running iOS 18.0+
Apple Intelligence-capable device for on-device AI extraction (iPhone 15 Pro or later)
Apple Intelligence enabled in Settings → Apple Intelligence & Siri

The app gracefully falls back to local parsing on all other devices.

Getting Started

1. Clone

git clone https://github.com/devendrabhumca12/ActionBrief.git
cd ActionBrief

2. Open in Xcode

open ActionBrief.xcodeproj

3. Configure signing

Select the ActionBrief target
Open Signing & Capabilities
Choose your Apple developer team
Update the bundle identifier if needed

4. Run

Choose an iPhone 15 Pro simulator or physical device
Press ⌘R to build and run
Tap Sample on the editor screen to load demo notes
Tap Extract Actions

5. Simulate AI availability states

Edit the scheme (Product → Scheme → Edit Scheme) and set Foundation Models Availability to simulate:

State	Behavior
Available	Full on-device AI extraction
Device not eligible	Falls back to local parser
Apple Intelligence disabled	Falls back with settings prompt
Model not ready	Shows "Model Loading" state

Project Structure

ActionBrief/
├── ActionBrief.xcodeproj/          # Xcode project (shared scheme included)
├── ActionBrief/
│   ├── ActionBriefApp.swift        # App entry point
│   ├── Models/
│   │   ├── Brief.swift             # Core data model
│   │   ├── ActionItem.swift        # Task model with owner/date/priority
│   │   ├── ActionPriority.swift    # Priority enum (low/medium/high)
│   │   ├── ExtractionResult.swift  # Service output wrapper
│   │   ├── AIAvailabilityState.swift
│   │   └── NoteImportError.swift
│   ├── Services/
│   │   ├── ActionExtractionService.swift        # Protocol
│   │   ├── FoundationModelExtractionService.swift # AI path
│   │   ├── FallbackExtractionService.swift      # Local parser
│   │   ├── ExtractionServiceFactory.swift
│   │   ├── AIAvailabilityService.swift
│   │   ├── BriefStore.swift                     # JSON persistence
│   │   ├── SpeechTranscriptionService.swift
│   │   ├── TranscriptFileImportService.swift
│   │   └── ImageTextExtractionService.swift     # Vision OCR
│   ├── ViewModels/
│   │   ├── BriefEditorViewModel.swift
│   │   ├── ActionReviewViewModel.swift
│   │   ├── BriefDetailViewModel.swift
│   │   └── SavedBriefsViewModel.swift
│   ├── Views/
│   │   ├── RootView.swift
│   │   ├── BriefEditorView.swift
│   │   ├── ActionReviewView.swift
│   │   ├── SavedBriefsView.swift
│   │   ├── BriefDetailView.swift
│   │   ├── TranscriptRecorderView.swift
│   │   ├── SettingsView.swift
│   │   └── TutorialView.swift
│   ├── Components/
│   │   ├── ActionItemRow.swift
│   │   ├── AvailabilityBanner.swift
│   │   ├── InputSourceBar.swift
│   │   ├── PrimaryButton.swift
│   │   ├── EmptyStateView.swift
│   │   └── LoadingStateView.swift
│   ├── Utilities/
│   │   ├── DateParsing.swift
│   │   ├── AppConstants.swift
│   │   ├── SampleData.swift
│   │   └── TutorialStorage.swift
│   └── Resources/
│       └── PrivacyInfo.xcprivacy   # Apple privacy manifest
├── docs/
│   ├── DEVELOPMENT.md              # Roadmap and release checklist
│   ├── PRIVACY.md                  # Full privacy policy
│   └── APP_STORE_REVIEW_NOTES.md
├── screenshots/
├── LICENSE
└── README.md

Privacy

ActionBrief is built privacy-first by design — not as an afterthought.

Notes stay on your device — zero network calls for note processing
No backend, account, or analytics — nothing to sign up for
No sensitive permissions in v1.0 — no contacts, location, or camera (Photos import uses system picker only)
Apple privacy manifest included at ActionBrief/Resources/PrivacyInfo.xcprivacy
App Store label: No Data Collected

Read the full policy in docs/PRIVACY.md.

Key Engineering Decisions

Why @Generable instead of raw prompt parsing? Structured generation with Foundation Models guarantees typed Swift output — no JSON extraction, no hallucinated field names, no runtime crashes from unexpected model output.

Why a fallback service? iOS 26 and Apple Intelligence are not universal. The protocol-based ActionExtractionService means the rest of the app has zero awareness of which path runs — the factory decides at startup, and the UI adapts via AIAvailabilityState.

Why local JSON storage instead of SwiftData? SwiftData requires iOS 17+ and adds migration complexity. For a v1.0 single-entity store, Codable + JSONEncoder is simpler, portable, and trivially testable. SwiftData is a clear upgrade path.

Roadmap

App Intent for clipboard extraction (Shortcuts integration)
Export to Reminders or Calendar (explicit user permission)
Priority grouping and search in saved briefs
Home Screen widgets for upcoming action items
Optional iCloud sync
Final app icon artwork
App Store release

See docs/DEVELOPMENT.md for the full checklist.

Contributing

Contributions are welcome! Please open an issue before submitting a pull request for anything beyond small bug fixes.

Fork the repository
Create a feature branch: git checkout -b feature/my-feature
Commit your changes: git commit -m 'Add my feature'
Push to the branch: git push origin feature/my-feature
Open a pull request

Why This Project

ActionBrief demonstrates practical on-device AI rather than a generic chatbot — a rarer and more interesting engineering story. It showcases:

Structured generation with Apple Foundation Models (@Generable)
Graceful degradation across the entire iOS 18–26 device spectrum
Privacy-first product thinking baked into architecture, not bolted on
Clean SwiftUI + MVVM + async/await with no third-party dependencies
Production-minded handling of error, loading, empty, and unavailable states

License

This project is licensed under the MIT License. See LICENSE.

Built by Devendra Kumar

If this project helps you, give it a ⭐

Navigation Menu

Repository files navigation

ActionBrief

Demo

Screenshots

The Problem It Solves

Features

Tech Stack

Architecture

Key Components

AI Workflow

Primary Path (iOS 26+, Apple Intelligence enabled)

Fallback Path (iOS 18–25 or AI disabled)

Requirements

Getting Started

1. Clone

2. Open in Xcode

3. Configure signing

4. Run

5. Simulate AI availability states

Project Structure

Privacy

Key Engineering Decisions

Roadmap

Contributing

Why This Project

License

About

Releases

Packages

Contributors

Languages

Name		Name	Last commit message	Last commit date
Latest commit History 5 Commits
ActionBrief.xcodeproj		ActionBrief.xcodeproj
ActionBrief		ActionBrief
docs		docs
screenshots		screenshots
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md

Folders and files

Latest commit

History

Repository files navigation

ActionBrief

Demo

Screenshots

The Problem It Solves

Features

Tech Stack

Architecture

Key Components

AI Workflow

Primary Path (iOS 26+, Apple Intelligence enabled)

Fallback Path (iOS 18–25 or AI disabled)

Requirements

Getting Started

1. Clone

2. Open in Xcode

3. Configure signing

4. Run

5. Simulate AI availability states

Project Structure

Privacy

Key Engineering Decisions

Roadmap

Contributing

Why This Project

License

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages