Newsworthy

A news anxiety dashboard. Scrapes headlines, runs them through GPT, and scores each category by the probability that a psychologist will have clients in tears about it in their next session. Categories scoring above 40% are shown prominently; lower-scoring ones are collapsed and can be expanded.

How it works

1. Headline scraping — fetch.py pulls headlines from 68k.news using curl and parses them with BeautifulSoup.
2. Scoring — The headlines are sent to GPT (currently gpt-5.2) with a prompt that asks it to score each news category (POLITICAL, GLOBAL, WAR / CONFLICT, ECONOMY, etc.) by the probability of causing client tears tomorrow, and to describe why people will be stressed about it.
3. JSON output — GPT returns a JSON blob with a [score, description] pair per category, plus an OVERALL score. This is written to news.json.
4. Refresh logic — news.json is considered stale after 5 minutes. A background process watches a FIFO; when the HTTP server detects a stale file, it signals the FIFO, and the background worker fetches fresh headlines and regenerates the JSON. The file is swapped atomically via rename.
5. HTTP server — A simple Python http.server serves index.html, style.css, script.js, and /news.json on port 8000.
6. Frontend — script.js fetches /news.json and renders:
   • A full-width summary banner at the top with the OVERALL score, a vibe label ("cancel your lunch break", "quiet day in the waiting room", etc.), the OVERALL description, and a clean underwear advisory scaled to severity.
   • Category tiles for anything scoring above 40%, colored green→red by score.
   • Collapsed <details> rows for lower-scoring categories, expandable on click.
   • The page auto-reloads every 2 minutes while the tab is focused.

Setup

Requirements: Python 3.9+, an OpenAI API key.

git clone <repo>
cd newsworthy

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

cp .env.example .env
# edit .env and set API_KEY=your_openai_api_key

Running locally

source venv/bin/activate
python fetch.py

Open http://localhost:8000.

On first run, news.json won't exist yet. The server will return an empty dashboard until the background worker finishes its first GPT request (usually 10–30 seconds).

Deploying to a VPS

A deploy.sh script sets up a systemd service and starts it:

# On the VPS, as a non-root user with sudo access:
./deploy.sh

The script will:

• Create a virtualenv and install dependencies
• Bootstrap an empty news.json so the app doesn't crash on first request
• Write and enable a systemd service that runs fetch.py and restarts on failure

The dashboard will be available at http://<server-ip>:8000.

Useful commands after deploying:

sudo journalctl -u newsworthy -f      # live logs
sudo systemctl restart newsworthy     # restart
sudo systemctl stop newsworthy        # stop

Files

File	Purpose
fetch.py	Scraping, GPT scoring, HTTP server, refresh logic
index.html	Page shell; triggers auto-reload every 2 minutes
script.js	Fetches /news.json, renders summary banner and tiles
style.css	Mobile-friendly layout, tile and banner styles
deploy.sh	Systemd deploy script for Linux VPS
.env.example	Environment variable template
requirements.txt	Python dependencies