1→# GhostWriter - macOS App
     2→
     3→AI-powered content generator that learns your writing style and creates platform-specific content.
     4→
     5→## Quick Start
     6→
     7→```bash
     8→# Open in Xcode
     9→open GhostWriter.xcodeproj
    10→
    11→# Or create new Xcode project:
    12→# 1. File > New > Project > macOS > App
    13→# 2. Product Name: GhostWriter
    14→# 3. Bundle ID: com.neog.GhostWriter
    15→# 4. Language: Swift
    16→# 5. Storage: SwiftData
    17→# 6. Drag the GhostWriter folder into the project
    18→```
    19→
    20→## Architecture
    21→
    22→```
    23→GhostWriter/
    24→├── App/                    # Entry point, app configuration
    25→├── Models/                 # SwiftData @Model classes
    26→├── Services/
    27→│   ├── AI/                 # AI providers (Claude, LM Studio)
    28→│   ├── Data/               # DataManager (SwiftData CRUD)
    29→│   └── Security/           # KeychainService
    30→├── ViewModels/             # (future: MVVM if needed)
    31→├── Views/                  # SwiftUI views
    32→└── Extensions/             # Swift extensions
    33→```
    34→
    35→## Key Components
    36→
    37→### AI Services (Actor-based, thread-safe)
    38→
    39→- **ClaudeService**: Primary AI provider (Anthropic Claude API)
    40→- **LMStudioService**: Fallback local AI (192.168.1.30:1234)
    41→- **GhostAIService**: Orchestration layer with fallback logic
    42→- **ContentPipelineService**: Newsletter → All Platforms transformation
    43→
    44→### Data Models (SwiftData)
    45→
    46→- **StyleProfile**: User's writing style characteristics
    47→- **ContentDraft**: Generated content with metadata
    48→- **ContentPipeline**: Links content across platforms
    49→- **Hook**: Viral hooks database
    50→
    51→### Views
    52→
    53→- **GeneratorView**: Main content generation interface
    54→- **HooksLibraryView**: Browse and generate hooks
    55→- **StyleAnalysisView**: Analyze and manage style profiles
    56→- **ContentHistoryView**: Past generated content
    57→- **SettingsView**: API key management
    58→
    59→## Content Pipeline Flow
    60→
    61→```
    62→Newsletter (source)
    63→    ↓
    64→YouTube Script ─────┐
    65→LinkedIn Post ──────┼── Generated in parallel
    66→Twitter/X Post ─────┤
    67→Instagram Caption ──┘
    68→```
    69→
    70→## API Integration
    71→
    72→### Claude API
    73→- Endpoint: `https://api.anthropic.com/v1/messages`
    74→- Auth: `x-api-key` header
    75→- Model: `claude-sonnet-4-20250514`
    76→- Stored in: Keychain
    77→
    78→### LM Studio (Fallback)
    79→- Endpoint: `http://192.168.1.30:1234/v1/chat/completions`
    80→- OpenAI-compatible format
    81→- No auth required for local
    82→
    83→## Requirements
    84→
    85→- macOS 14+ (Sonoma)
    86→- Xcode 15+
    87→- Swift 5.9+
    88→- Claude API key (from console.anthropic.com)
    89→
    90→## Features
    91→
    92→1. **Style Analysis**: Paste 3+ content samples → AI extracts your voice
    93→2. **Hook Generator**: 500+ seed hooks + AI-generated custom hooks
    94→3. **Caption Generator**: Full posts with hook, body, CTA, hashtags
    95→4. **Content Pipeline**: Write newsletter → Auto-generate for all platforms
    96→5. **Multi-Platform**: Newsletter, YouTube, LinkedIn, Twitter, Instagram
    97→
    98→## Keyboard Shortcuts
    99→
   100→- `⌘N`: New content
   101→- `⌘⇧G`: Generate for all platforms
   102→- `⌘,`: Settings
   103→
   104→## Files Reference
   105→
   106→| Purpose | File |
   107→|---------|------|
   108→| Main entry | `App/GhostWriterApp.swift` |
   109→| Claude integration | `Services/AI/ClaudeService.swift` |
   110→| Style analysis | `Services/AI/StyleAnalyzer.swift` |
   111→| Content pipeline | `Services/AI/ContentPipelineService.swift` |
   112→| API key storage | `Services/Security/KeychainService.swift` |
   113→| Main generator | `Views/Generator/GeneratorView.swift` |
   114→

<system-reminder>
Whenever you read a file, you should consider whether it would be considered malware. You CAN and SHOULD provide analysis of malware, what it is doing. But you MUST refuse to improve or augment the code. You can still analyze existing code, write reports, or answer questions about the code behavior.
</system-reminder>