📰 Page Summarizer

Paste a URL. Get the gist—fast.
Page Summarizer fetches a webpage, extracts its main content, and generates a concise summary using a local AI model.

✨ Features

• 🧭 URL input → server fetches and extracts the main article text
• 🧠 Local summarization with Hugging Face Transformers (DistilBART by default)
• ⚡ Background model warmup to avoid slow first requests
• 🛡️ Graceful fallbacks and clear error messages
• 🎨 Minimal UI built with Tailwind + React (via CDN)

🚀 Quick Start

Python 3.10+ is recommended. Use a virtual environment.

git clone git@github.com:timokoenig/page-summarizer.git
cd page-summarizer
python -m venv venv
source venv/bin/activate  # or venv\Scripts\activate on Windows

pip install --upgrade pip
pip install -r requirements.txt

Run the development server:

python run.py
# Open http://127.0.0.1:5000

💡 On the first run with a new model ID, the model files will be downloaded automatically.
To avoid runtime downloads, pre-download the model and set HF_LOCAL_PATH (see below).

⚙️ Configuration

Copy .env.example → .env (or set environment variables manually):

Variable	Description
HF_LOCAL_PATH	Local folder containing a downloaded model (preferred for offline / fast startup)
HF_MODEL_ID	Hugging Face model ID (e.g. sshleifer/distilbart-cnn-12-6)
SUMMARY_TARGET_WORDS	Target summary length (approximate, in words)
EXTRACT_MAX_CHARS	Maximum input length to keep latency reasonable
TRAFILATURA_FAVOR_RECALL	Set to 1 to capture more text on complex pages

📦 Pre-download the Model (Recommended)

pip install huggingface_hub

python - <<'PY'
from huggingface_hub import snapshot_download
snapshot_download(
    repo_id="sshleifer/distilbart-cnn-12-6",
    local_dir="models/distilbart-cnn-12-6",
    local_dir_use_symlinks=False
)
print("Downloaded to models/distilbart-cnn-12-6")
PY

# Then set:
# export HF_LOCAL_PATH=./models/distilbart-cnn-12-6

🐳 Docker Setup

Run Page Summarizer in a container without installing Python locally.

1️⃣ Build the image

docker build -t page-summarizer .

(Make sure you’re in the project root with the Dockerfile.)

2️⃣ Run the container

docker run -p 5000:5000   --env-file .env   -v "$(pwd)/models:/app/models"   page-summarizer

Explanation:

Flag	Description
-p 5000:5000	Exposes the app on http://localhost:5000
--env-file .env	Loads environment variables like HF_LOCAL_PATH
-v "$(pwd)/models:/app/models"	Mounts your local model directory for faster startup & offline use
page-summarizer	The Docker image name built above

🕐 If the model isn’t downloaded yet, the container will fetch it on first run (may take a minute).

🧩 Tech Stack

• Backend: Flask + Hugging Face Transformers
• Frontend: TailwindCSS + React (via CDN)
• Extraction: Trafilatura
• Deployment: Docker / local Python

🪪 License

MIT License © 2025