LangGraph Tutorial with ZenRows

A LangGraph pipeline that scrapes news articles with ZenRows, extracts structured fields with an LLM, and synthesizes a short briefing.

How it works

The graph (graph.py) wires together four nodes (nodes.py):

1. scrape — fetches each URL with ZenRowsUniversalScraper (JS rendering + premium proxy).
2. retry — if a scrape looks blocked (bot-check page, empty response), retries once with a US proxy.
3. extract — asks gpt-4o-mini to pull headline, source, date, and summary from each article as JSON.
4. synthesize — asks the LLM to turn the extracted articles into a short bullet-point briefing.

scrape --(blocked?)--> retry --> extract --> synthesize --> END
   \_________________(ok)_______/

Setup

pip install -r requirements.txt

Create a .env file in the project root with:

ZENROWS_API_KEY=your_zenrows_api_key
OPENAI_API_KEY=your_openai_api_key

Run

python graph.py

This scrapes a hardcoded list of URLs (edit the urls list in graph.py to change targets), prints the extracted fields, and prints the final briefing.