NewsSummarizer reads article URLs from an Excel file, scrapes the article content, generates a consolidated AI news brief using a local Ollama model, and emails the final summary.

• Reads news links from NewsLinks.xlsx
• Scrapes article content using crawl4ai
• Summarizes all stories via local Ollama model (llama3.2:3b by default)
• Produces a markdown-style newsletter summary with inline source links
• Sends the summary by email (plain text + HTML)
• Falls back to printing summary in console if email credentials are missing

• Python 3.10+
• Ollama installed and running locally
• Chromium browser support for crawler runtime (playwright install)

Install required packages:

pip install langgraph langchain_ollama langchain-core crawl4ai pandas openpyxl python-dotenv

Install Playwright browser dependencies:

playwright install

Start Ollama and pull the model used by default:

ollama pull llama3.2:3b

You can override the model with OLLAMA_MODEL in .env.

Create a .env file in project root:

EMAIL_ADDRESS=your_email@gmail.com
EMAIL_PASSWORD=your_gmail_app_password
RECIPIENT_EMAIL=recipient@gmail.com
OLLAMA_MODEL=llama3.2:3b

Notes:

• EMAIL_PASSWORD should be a Gmail App Password, not your account password.
• If email variables are missing, the script prints the summary to console instead of sending mail.

The script expects an Excel file named NewsLinks.xlsx in project root with a column named exactly:

• URL

Example:

Pipeline implemented in Agent.py:

1. Read URLs from NewsLinks.xlsx (read_excel_node)
2. Scrape each URL and extract text (scrape_links_node)
3. Summarize all scraped content with Ollama (summarize_node)
4. Email summary as newsletter (send_email_node)

The flow is orchestrated using LangGraph:

START -> read_excel -> scrape -> summarize -> email -> END

From project root:

python Agent.py

• Success path: summary is emailed to RECIPIENT_EMAIL
• Partial scrape failures: failed URLs are skipped; successful articles are still summarized
• No scrape success: execution stops with error
• Email send failure: summary is printed in terminal

• ModuleNotFoundError: install missing package(s) with pip install ...
• Playwright/browser issues: run playwright install
• Ollama connection/model errors:
  • ensure Ollama daemon is running
  • ensure model exists (ollama pull llama3.2:3b)
• Excel errors:
  • verify file is named NewsLinks.xlsx
  • verify URL column exists and contains valid links
• Gmail auth errors:
  • confirm App Password is used
  • confirm account allows app password usage

• Do not commit .env to source control
• Rotate credentials immediately if exposed
• Use a dedicated sender mailbox for automation

pip install -r requirements.txt

uvicorn app:app --reload