Daily AI News Digest Agent — README

Overview

This Python script automates the process of collecting, filtering, summarizing, and emailing a daily AI/ML news digest. It gathers articles from multiple AI-related RSS feeds, filters items published within the last 24 hours, removes duplicates, summarizes them using Google Gemini, and sends a formatted Hebrew-language digest directly to your inbox.

Features

• Fetches AI/ML news from curated RSS feeds (OpenAI, DeepMind, Hugging Face, arXiv, TechCrunch AI, The Verge AI, etc.).
• Extracts only articles published in the last 24 hours.
• Removes duplicate news items across feeds using URL-based deduplication.
• Uses Google Gemini to generate a clean, structured Hebrew summary.
• Developer-focused digest: SOTA models, tools, new releases, frameworks, major updates.
• Sends an HTML email with RTL support, custom styling, and timestamp.
• Fully automated — can be run locally, via cron, or through GitHub Actions.

Requirements

• Python 3.9+ (3.10 or 3.11 recommended)
• Internet access for RSS fetching + Gemini API calls
• Valid Gemini API Key
• Email account with SMTP access (Gmail recommended)

Installation

1. Clone the repository git clone https://github.com/your-repo/daily-ai-digest.git cd daily-ai-digest
2. Install required Python packages Required packages include:

• requests
• feedparser
• google-generativeai
• smtplib (standard library)
• email (standard library)

Environment Variables

Set these environment variables before running the script:

• GEMINI_API_KEY Your Google Gemini API key
• EMAIL_SENDER Email address used to send messages
• EMAIL_PASSWORD App password or SMTP password
• EMAIL_RECEIVER Target email address to receive digest

How It Works

1. Fetch News (fetch_news)

• Downloads RSS feeds using browser-like headers (prevents blocking from Meta/Google).
• Parses publication time from each entry.
• Keeps only items from the last 24 hours.
• Removes duplicates based on link.
• Outputs a large aggregated text block used for summarization.

2. Summarize with Gemini (summarize_news)

• Sends the aggregated raw text to Gemini (multiple fallback models).
• Generates a Hebrew developer-focused digest.
• Ensures: No intros, no outros. Each news item includes headline, date, summary, details, link. Items are not merged unless identical.

3. Email the Digest (send_email)

• Converts the LLM output to simple HTML.
• Adds a styled header with RTL alignment.
• Sends using Gmail SMTP (smtp.gmail.com:587).

Running the Script

python main.py

If no news was found in the last 24 hours, it prints: No news found.

Automation (Optional) — GitHub Actions

You can fully automate the daily digest using GitHub Actions.

1. Configure Secrets Inside your GitHub repository:

• Go to: Settings → Secrets and variables → Actions
• Click New repository secret
• Add the following secrets:

GEMINI_API_KEY Your Gemini API key EMAIL_SENDER Email used to send messages EMAIL_PASSWORD App password (recommended) EMAIL_RECEIVER Recipient email

2. Create the Workflow

• Go to the Actions tab
• Click set up a workflow yourself
• Name the file: daily_run.yml
• Paste the following:

name: Daily AI News Aggregator

on:
  schedule:
    # Runs every day at 06:00 UTC (08:00–09:00 Israel time depending on DST)
    - cron: '0 6 * * *'
  workflow_dispatch: # Allows manual runs for testing

jobs:
  build:
    runs-on: ubuntu-latest

    steps:
    - name: Checkout code
      uses: actions/checkout@v3

    - name: Set up Python
      uses: actions/setup-python@v4
      with:
        python-version: '3.9'

    - name: Install dependencies
      run: |
        pip install -r requirements.txt

    - name: Run script
      env:
        GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}
        EMAIL_SENDER: ${{ secrets.EMAIL_SENDER }}
        EMAIL_PASSWORD: ${{ secrets.EMAIL_PASSWORD }}
        EMAIL_RECEIVER: ${{ secrets.EMAIL_RECEIVER }}
      run: python daily_ai_agent.py

• Click Commit changes

Your digest will now be generated and emailed automatically every day.

Troubleshooting

• 403 errors on RSS feeds — Some sources block bots; custom User-Agent solves this.
• Email not sending — Gmail often requires an app password if 2FA is enabled.
• Gemini model errors — The script automatically retries multiple model versions.

Enjoy :)