A full-stack web app that curates arXiv papers, generates concise AI summaries, ranks relevance, and delivers a personalized daily digest.

• Topic Onboarding — Select from curated research topics (LLMs, ML, CV, Robotics, etc.) or create custom ones
• arXiv Paper Ingestion — Fetches recent papers from arXiv based on your topics, with deduplication
• AI Summarization — Each paper gets a 3–5 sentence summary, key contributions, practical relevance, limitations, and importance score (powered by OpenAI, Anthropic, or a built-in mock)
• Ranking / Scoring — Papers are ranked 0–100 using recency, novelty keywords, topic match, and optional LLM scoring
• Daily Digest — Top 5–10 papers curated just for you, with estimated reading time
• Paper Feed — Browse all papers with search, topic filter, and score filter
• Paper Detail — Full metadata, AI summary, related papers, save/read actions
• Save & Read — Mark papers as saved for later or already read
• Scheduled Sync — Background job refreshes papers every N hours

PaperPulse/
├── backend/
│   ├── app/
│   │   ├── api/            # FastAPI route handlers
│   │   ├── services/       # Business logic (arXiv, LLM, ranking, digest)
│   │   ├── models/         # SQLAlchemy ORM models
│   │   ├── schemas/        # Pydantic request/response schemas
│   │   ├── jobs/           # APScheduler background sync
│   │   ├── utils/          # Seed data
│   │   ├── config.py       # Settings (pydantic-settings)
│   │   ├── database.py     # DB engine & session
│   │   └── main.py         # FastAPI app entry point
│   ├── seed.py             # DB seed script
│   ├── requirements.txt
│   └── .env.example
│
└── frontend/
    ├── app/
    │   ├── page.tsx          # Research feed
    │   ├── digest/page.tsx   # Daily digest
    │   ├── topics/page.tsx   # Topic management
    │   └── papers/[id]/page.tsx  # Paper detail
    ├── components/
    │   ├── PaperCard.tsx
    │   ├── Providers.tsx
    │   ├── layout/Sidebar.tsx
    │   └── ui/
    ├── lib/
    │   ├── api.ts            # API client
    │   ├── types.ts          # TypeScript types
    │   └── utils.ts
    └── .env.local.example

• Python 3.11+
• Node.js 18+

cd backend

# Create virtual environment
python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Configure environment
cp .env.example .env
# Edit .env — set LLM_PROVIDER if desired (default: mock)

# Seed database and default topics
python seed.py

# (Optional) fetch demo papers from arXiv
python seed.py --fetch

# Start the backend server
uvicorn app.main:app --reload --port 8000

The API will be available at http://localhost:8000.
Interactive docs: http://localhost:8000/docs

cd frontend

# Install dependencies
npm install

# Configure environment
cp .env.local.example .env.local

# Start the development server
npm run dev

Open http://localhost:3000 in your browser.

After starting the backend, click Sync Papers in the sidebar, or trigger it via the API:

curl -X POST http://localhost:8000/api/sync/

Papers will appear in the feed within seconds (with the mock LLM provider).

No API key needed. Returns deterministic summaries based on paper metadata. Great for local development and demos.

LLM_PROVIDER=openai
OPENAI_API_KEY=sk-...

Uses gpt-4o-mini for cost-effective summarization.

LLM_PROVIDER=anthropic
ANTHROPIC_API_KEY=sk-ant-...

Uses claude-3-haiku-20240307.

For production or persistent storage, switch to PostgreSQL:

DATABASE_URL=postgresql://user:password@localhost:5432/paperpulse

Install the driver:

pip install psycopg2-binary

• The backend uses auth-free single-user mode (User ID = 1). Multi-user support can be added by extending the auth layer.
• Summaries are cached in the database — they won't be regenerated on subsequent syncs.
• The LLM layer is fully pluggable: implement the LLMProvider protocol in app/services/llm_service.py to add new providers.
• The scheduler runs in-process (APScheduler). For production, consider using Celery or a cron job.