Daily News Report Agent
Goal: Generate high-quality, curated news digests from various internet sources (RSS, APIs, Search).
- Information Gathering
-
Sources:
-
RSS Feeds: TechCrunch, Hacker News, Verges, specialized industry blogs.
-
APIs: NewsAPI, GNews, Reddit API.
-
Direct Scrape: Use with caution and respect robots.txt (using firecrawl-scraper ).
-
Filtering:
-
Keywords: Filter by relevant topics (e.g., "AI", "Rust", "Global Economy").
-
Recency: Only fetch items < 24h.
-
Popularity: Filter by score/upvotes/comments if available (e.g., HN > 100 points).
- Processing Pipeline
-
Deduplication: Use similarity check (Title/Content embeddings) to group same stories from different sources.
-
Summarization: Use LLM to summarize content into 3-5 bullet points. Focus on "Why it matters".
-
Classification: Tag articles with categories (Tech, Business, Politics, Science).
- Reporting Output
-
Format: Markdown, HTML email, or Slack/Discord payload.
-
Structure:
-
Headline: Catchy but accurate.
-
TL;DR: 1 sentence summary.
-
Key Points: Bulleted list.
-
Source Link: Direct link to original article.
-
Sentiment: Neutral/Positive/Negative tag (optional).
- Scheduled Automation
-
Trigger: Cron job (e.g., every morning at 8:00 AM).
-
Environment: GitHub Actions (Scheduled workflow) or a persistent background worker (Node.js/Python).
- Example Workflow (GitHub Action)
name: Daily News on: schedule: - cron: '0 8 * * *' jobs: report: runs-on: ubuntu-latest steps: - uses: actions/checkout@v3 - name: Run News Agent run: npx antigravity-ide run-skill daily-news-report env: NEWS_API_KEY: ${{ secrets.NEWS_API_KEY }} LLM_API_KEY: ${{ secrets.GEMINI_API_KEY }}
Prompt Strategy for Summarization:
"You are a professional news editor. Summarize the following text for a technical audience. Focus on facts, release numbers, and technical implications. Remove fluff."