📄 PaperBlog

A modern blog that automatically fetches, summarizes, and publishes the top AI research papers from Hugging Face Daily Papers every day.

Features

• 🤖 Daily auto-fetch — Scrapes top-N papers from Hugging Face (not RSS, actual web scraping)
• 📝 AI summaries — Generates plain-language summaries via OpenAI GPT-4o-mini or HF Inference API
• 🌐 Modern blog UI — Built with Next.js 14 (App Router) + Tailwind CSS
• 🔌 MCP Agent API — MCP-compatible REST endpoint at /api/mcp for AI agent integration
• 📦 REST API — Full REST API at /api/papers for external integrations
• 📅 Archive — Browse all historical daily paper batches
• ⚡ GitHub Actions — Fully automated daily workflow

Quick Start

Prerequisites

• Node.js 22+
• Python 3.12+

1. Install dependencies

# Frontend
npm install

# Python scripts
pip install -r requirements.txt

2. Fetch today's papers

# Basic fetch (no AI summaries)
python scripts/fetch_papers.py

# With AI summaries (requires OPENAI_API_KEY or HF_API_KEY)
OPENAI_API_KEY=sk-... python scripts/fetch_papers.py --summarize

# Specific date, top 5 papers
python scripts/fetch_papers.py --date 2025-02-28 --top 5 --summarize

3. Run the blog locally

npm run dev
# Open http://localhost:3000

Configuration

Environment Variables

Variable	Description
OPENAI_API_KEY	OpenAI API key for GPT-4o-mini summaries
HF_API_KEY	Hugging Face API key for BART summarization
TOP_N	Number of papers to fetch per day (default: 10)
OUTPUT_DIR	Directory for paper JSON files (default: content/papers)

GitHub Actions Setup

1. Add secrets in your GitHub repo settings:

   • OPENAI_API_KEY (optional, for AI summaries)
   • HF_API_KEY (optional, for HF-based summaries)

2. The workflow runs automatically at 08:00 UTC daily.

3. Trigger manually via Actions → Daily Papers → Run workflow.

API Reference

GET /api/papers

List papers. Query params: date=YYYY-MM-DD, q=<search>, limit=N

GET /api/papers/:id

Get a single paper by arXiv ID.

GET /api/mcp

MCP server manifest — returns available tools.

POST /api/mcp

Execute an MCP tool call.

{
  "name": "get_latest_papers",
  "arguments": { "limit": 5 }
}

Available MCP Tools

Tool	Description
get_latest_papers	Get the most recent daily papers
get_papers_by_date	Get papers for a specific date
get_paper_by_id	Get a single paper by arXiv ID
search_papers	Full-text search across all papers
list_dates	List all available dates

Architecture

PaperBlog/
├── src/
│   ├── app/                    # Next.js App Router
│   │   ├── page.tsx            # Homepage (latest papers)
│   │   ├── archive/            # Archive page
│   │   ├── papers/[slug]/      # Paper detail page
│   │   └── api/
│   │       ├── papers/         # REST API
│   │       └── mcp/            # MCP agent endpoint
│   ├── components/             # React UI components
│   ├── lib/                    # Data access layer
│   └── types/                  # TypeScript types
├── scripts/
│   ├── fetch_papers.py         # HF scraper + summarizer
│   └── summarize.py            # Re-run summarization on existing data
├── content/
│   └── papers/                 # Daily paper JSON files (YYYY-MM-DD.json)
└── .github/workflows/
    └── daily-papers.yml        # GitHub Actions automation

Extending

• New AI providers: Add a summarize_with_* function in scripts/fetch_papers.py
• Electron/Desktop app: Consume /api/papers endpoints
• Angular frontend: Replace or augment src/ with an Angular app consuming the API
• MCP agents: Point your MCP client at POST /api/mcp with tool calls

License

MIT