An AI-powered study assistant that transforms PDFs into personalized learning experiences. Upload a document, pick your mood, and Clarify generates flashcards, cramming sessions, or summaries — all powered by Claude.

flowchart TD
    User(["👤 User"])

    subgraph Client["Browser (Next.js Client Components)"]
        Home["/ Home Page"]
        FC["📚 /flashcards\nCreate & manage decks"]
        CR["🧠 /cramming\nInteractive study session"]
        SM["📄 /summary\nAI document summary"]
        LS[("💾 localStorage\nFlashcard decks")]
    end

    subgraph Server["Next.js API Routes (Server)"]
        EX["/api/extract\nPDF → Text\n(unpdf)"]
        FL["/api/flashcards\nText → Flashcards\n(Claude Opus 4.6)"]
        CF["/api/chat-flashcards\nConversation → Flashcards\n(Claude Opus 4.6)"]
        SU["/api/summarize\nText → Summary\n(Claude Opus 4.6)"]
        AT["/api/analyze-topics\nWrong cards → Weak topics\n(Claude Haiku 4.5)"]
    end

    subgraph Anthropic["Anthropic API"]
        OP["Claude Opus 4.6\nFlashcard & Summary generation"]
        HK["Claude Haiku 4.5\nWeak topic analysis"]
    end

    PDF[["📎 PDF File"]]

    User --> Home
    Home --> FC & CR & SM

    FC -->|"Upload PDF"| PDF
    FC -->|"Chat with AI"| CF
    SM -->|"Upload PDF"| PDF

    PDF -->|"POST /api/extract"| EX
    EX -->|"Extracted text"| FL
    EX -->|"Extracted text"| SU

    FL -->|"POST /api/flashcards"| OP
    CF -->|"POST /api/chat-flashcards"| OP
    SU -->|"POST /api/summarize"| OP
    AT -->|"POST /api/analyze-topics"| HK

    OP -->|"Flashcard JSON"| FC
    OP -->|"Flashcard JSON"| CF
    OP -->|"Summary markdown"| SM
    HK -->|"Topic strings"| CR

    FC -->|"Save decks"| LS
    LS -->|"Load decks"| CR

    CR -->|"Session complete\n(wrong cards)"| AT

• Upload any PDF and generate flashcards instantly

• Choose a mood to control depth and quantity:

  Mood	Cards	Style
  😴 Tired	5	Short & simple
  😰 Stressed	3	Critical points only
  😤 Annoyed	5	Blunt, no fluff
  🤓 Curious	10	Deep & detailed

• Name decks, expand to preview cards inline, edit individual Q&A pairs

• Add and delete cards manually

• All decks persisted in localStorage

• Generate flashcards on any topic through natural conversation — no PDF required
• Toggle "Chat with AI" on the flashcards page to open the chat panel
• Describe a topic and Claude generates 8 flashcards by default; ask for more, fewer, or a different focus and it regenerates the full set
• Claude may ask one clarifying question (depth, level, specific focus) before generating

• Pick any saved deck and start studying immediately
• Flip cards by clicking or keyboard shortcuts
• Mark each card Know It or Don't Know
• Real-time progress bar showing known / unknown / unreviewed counts
• Session statistics on completion:
  • Accuracy percentage and performance grade
  • Time spent total and per card
  • AI-identified weak topics from missed cards
  • Full list of cards you got wrong
• Retry only the wrong cards, or reset and go again
• Rename the deck inline during a session
• End session early with "Finish Session" button

• Upload a PDF and receive a mood-tailored summary:

  Mood	Style
  😴 Tired	5 bullets, ≤ 10 words each + funny analogy
  😰 Stressed	3 numbered critical points, calm tone
  😤 Annoyed	Blunt, no intro, fewest words possible
  🤓 Curious	Deep dive with insights and real-world context

• Two-stage loading: shows "Extracting PDF..." then "Generating summary..."

• Rendered markdown with styled headings, bullets, and inline bold

• Copy to clipboard or download as a formatted PDF

The app's core innovation is the mood system — four distinct emotional states that shape how the AI generates content. Each mood controls both the quantity and style of output.

Choose your current mental state when generating content:

clarify/
├── app/
│   ├── api/
│   │   ├── extract/route.js            # PDF → plain text (unpdf)
│   │   ├── flashcards/route.js         # Text → flashcard JSON (Claude Opus)
│   │   ├── chat-flashcards/route.js   # Conversation → flashcards (Claude Opus)
│   │   ├── summarize/route.js        # Text → summary markdown (Claude Opus)
│   │   └── analyze-topics/route.js    # Wrong cards → weak topics (Claude Haiku)
│   ├── cramming/
│   │   └── page.tsx                   # Interactive study session
│   ├── flashcards/
│   │   └── page.tsx                   # Deck management + creation + AI chat
│   ├── summary/
│   │   └── page.tsx                   # PDF summarization
│   ├── utils/
│   │   ├── aiApi.ts                   # Client-side API helpers + types
│   │   ├── crammingHelpers.ts         # Session logic helpers
│   │   └── deckAccents.ts             # Deck color accent classes
│   ├── globals.css
│   ├── layout.tsx
│   └── page.tsx                       # Landing page
├── components/
│   ├── flashcards/                    # Flashcard components
│   │   ├── CreateDeckView.tsx
│   │   ├── EditDeckView.tsx
│   │   ├── DeckCard.tsx
│   │   ├── ChatDeckCreator.tsx
│   │   ├── QACard.tsx
│   │   ├── ModeToggle.tsx
│   │   ├── DeckActions.tsx
│   │   └── ConfirmDeleteModal.tsx
│   ├── summary/                       # Summary components
│   │   ├── FileUpload.tsx
│   │   ├── SummaryContent.tsx
│   │   ├── RenderSummary.tsx
│   │   ├── ProcessingPDF.tsx
│   │   ├── MoodSelector.tsx
│   │   └── ActionsSection.tsx
│   └── ui/                            # shadcn/ui base components
│       ├── button.tsx
│       └── navigation-menu.tsx
├── lib/
│   ├── utils.ts                       # cn() className merger
│   └── prompts.js                     # Mood-based prompt templates
├── utils/
│   ├── flashcardStorage.ts            # localStorage helpers
│   └── pdfExport.ts                   # PDF download functionality
├── package.json
├── next.config.ts
└── tsconfig.json

• Node.js 18+
• An Anthropic API key

git clone https://github.com/SamuelIVX/clarify.git
cd clarify
npm install

Create a .env.local file in the project root:

# .env.local
ANTHROPIC_API_KEY=your_anthropic_api_key_here

npm run dev

Open http://localhost:3000.

npm run build
npm start

The app is deployed on Vercel. To deploy your own:

1. Push to main — Vercel auto-deploys on every push when connected to GitHub.
2. Add ANTHROPIC_API_KEY in your Vercel project under Settings → Environment Variables.

The /api/extract route sets maxDuration = 60 to handle large PDFs within Vercel's serverless timeout.

All routes run server-side only. ANTHROPIC_API_KEY is never exposed to the client.