RecallFlow AI

Generate AI-powered flashcards from your Obsidian notes, then review them using the FSRS (Free Spaced Repetition Scheduler) algorithm — all locally within your vault.

✨ Features

🧠 AI Flashcard Generation

• From selection — highlight text and generate a single focused card via right-click or command palette
• From entire note — process the full Markdown document into 3–6 mixed-type flashcards
• Preview & refine — inspect, edit, and regenerate AI-generated cards before saving
• Three card types: Q&A, Cloze (fill-in-the-blank), Multiple Choice

📆 FSRS Spaced Repetition

• Cards follow FSRS across 4 lifecycle states: New → Learning (~10min–1d) → Review (e.g., 3d ➔ 10d ➔ 1mo) → Relearning (≤1d)
• Interactive review: typed answers for Q&A, fill-the-blank for cloze, tap-to-select for choice
• Four ratings: Again (1), Hard (2), Good (3), Easy (4) to fine-tune memory curve
• Configure maximum new and review cards per day

🔥 Cramming Mode (Rerun)

• Re-quiz today's reviewed cards without affecting FSRS scheduling
• Full state persistence — navigate freely without losing answers

📋 Card Manager

• Search, edit, validate, and delete flashcards
• View metadata: card type, FSRS state, last review, due date, source, creation time

📊 Activity Heatmap

• 12-month GitHub-style contribution grid of daily review counts

🌐 Multi-Language UI

• English, 中文, 日本語, 한국어
• AI-generated card language configurable independently

🔒 Privacy-First

• No telemetry, no background network requests
• All data stored locally in vault plugin data
• Review works entirely offline

📖 How It Works

1. Configure your LLM endpoint (base URL, API key, model) in Settings → RecallFlow AI
2. Generate cards from selection or entire note via commands, ribbon, or right-click
3. Preview in the generation modal — edit, regenerate, then confirm to save
4. Review due cards in the immersive modal — answer, rate, repeat
5. Rerun to cram today's cards without resetting the FSRS algorithm
6. Manage your card library with search, edit, and validation tools

🚀 Installation

⚙️ Settings

🎯 Commands

Right-click context menu also offers generation when text is selected.

💬 Card Types

All three types are generated together. The AI decides the mix based on your note content.

🔧 Development

Requirements: Node.js 18+, npm

npm install
npm run dev       # Watch mode — writes main.js on change
npm run build     # Production bundle
npm run lint      # ESLint check

Copy main.js, manifest.json, styles.css into <vault>/.obsidian/plugins/recallflow-ai/ for local testing.

Project Structure

src/
├── main.ts                    Entry point, commands, ribbon, menus
├── types.ts                   Type definitions, defaults
├── services/
│   ├── fsrsService.ts         FSRS-5 scheduler, due-card queue, limits
│   ├── i18n.ts                Translations (en / zh / ja / ko)
│   └── llmService.ts          LLM prompt, API call, parser
└── ui/
    ├── cardManageModal.ts     Browser, search, edit, delete
    ├── cardPreviewModal.ts    AI preview with editing & regenerate
    ├── reviewModal.ts         Interactive review, ratings, cramming
    └── settingsTab.ts         Settings, test connection, heatmap

Tech Stack

🔒 Privacy & Data

• No telemetry — zero background network requests
• Generation data: note text + model name → your configured LLM endpoint only
• All data local — cards, history, stats, settings stored in vault plugin data by Obsidian

📄 License

0BSD

🤝 Contributing

See CONTRIBUTING.md. Open an issue for large changes, submit PRs that pass npm run build && npm run lint.

📚 Releasing

See RELEASING.md for the release checklist and Community directory submission process.