OncoPulse

OncoPulse — Weekly Oncology Literature Dashboard

Automated weekly digest of oncology literature: Phase 2/3 trials, FDA approvals, and key reviews — summarized by Claude and served as a searchable web dashboard.

How It Works

PubMed API ──┐
FDA RSS     ──┼──► fetch_and_summarize.py ──► data.json ──► index.html (dashboard)
Journal RSS ──┘          (Claude API)

Every Monday at 6:00 AM UTC, GitHub Actions:

Pulls recent articles from PubMed, FDA, and 10 major journal RSS feeds
Filters for clinical relevance (Phase 2+, FDA actions, reviews)
Calls Claude to generate structured summaries (design, endpoints, results, clinical takeaway)
Commits updated data.json to the repo
GitHub Pages serves the dashboard automatically

Setup

1. Fork / clone this repo

git clone https://github.com/YOURUSERNAME/oncopulse.git
cd oncopulse

2. Add secrets to GitHub

Go to Settings → Secrets and variables → Actions → New repository secret:

Secret	Value	Required
`ANTHROPIC_API_KEY`	Your Anthropic API key from console.anthropic.com	✅ Yes
`PUBMED_API_KEY`	From ncbi.nlm.nih.gov/account (free) — raises rate limit from 3 to 10 req/sec	Optional

3. Enable GitHub Pages

Go to Settings → Pages:

Source: Deploy from a branch
Branch: main / folder: /dashboard
Save

Your dashboard will be live at: https://YOURUSERNAME.github.io/oncopulse/

4. Run the first update manually

Go to Actions → OncoPulse Weekly Literature Update → Run workflow

Or run locally:

cd pipeline
pip install -r requirements.txt

export ANTHROPIC_API_KEY="sk-ant-..."
export OUTPUT_DIR="../dashboard"

python fetch_and_summarize.py --days 7 --max-articles 60

Configuration

Edit environment variables in .github/workflows/weekly_update.yml or set locally:

Variable	Default	Description
`LOOKBACK_DAYS`	`7`	Days of literature to fetch
`MAX_ARTICLES`	`60`	Max articles to summarize per run (cost control)
`OUTPUT_DIR`	`../dashboard`	Where to write `data.json`
`CACHE_DIR`	`.cache`	Cache directory (avoids re-summarizing seen articles)

Cron schedule

Edit .github/workflows/weekly_update.yml:

- cron: "0 6 * * 1"   # Monday 6 AM UTC — change to suit your timezone

Cost Estimate

Using Claude Sonnet at ~$3/MTok input + $15/MTok output:

Articles/week	Est. tokens	Est. cost
30	~120K	~$0.50
60	~240K	~$1.00
100	~400K	~$1.65

The .cache/ directory prevents re-summarizing articles already seen, keeping costs low on re-runs.

Project Structure

oncopulse/
├── pipeline/
│   ├── fetch_and_summarize.py   # Main pipeline script
│   ├── requirements.txt
│   └── .cache/                  # Auto-created, cached summaries
├── dashboard/
│   ├── index.html               # The web dashboard
│   └── data.json                # Generated weekly by pipeline
├── .github/
│   └── workflows/
│       └── weekly_update.yml    # GitHub Actions schedule
└── README.md

Sources

Source	What it pulls
PubMed	Phase 2/3 RCTs, meta-analyses published in last 7 days
FDA RSS	FDA oncology drug approvals and regulatory actions
NEJM	New England Journal of Medicine
Lancet Oncol	The Lancet Oncology
JCO	Journal of Clinical Oncology
Blood	ASH Blood journal
Cancer Cell	Cell Press Cancer Cell
Nature Cancer	Nature Cancer
JAMA Oncol	JAMA Oncology
Ann Oncol	Annals of Oncology
Cancer Discov	Cancer Discovery (AACR)

Local Development

Open dashboard/index.html in a browser. If no data.json exists yet, you’ll see a helpful error message. Run the pipeline first to generate it.

For local testing with a small sample:

python fetch_and_summarize.py --days 3 --max-articles 10 --skip-pubmed