Spectrum - AI News Summarizer

Automated AI news agent that monitors RSS feeds, summarizes insights using Perplexity Sonar, and delivers daily digests to Telegram.

Overview

A Python tool that runs locally on your machine. When executed, it fetches AI news from multiple RSS feeds, generates concise summaries using the Perplexity API, and posts them to a Telegram channel/bot. The tool includes local storage for all fetched articles and summaries, plus a final overview summary that consolidates all news into a single digest.

Key Features:

• 📰 Monitors multiple AI news RSS feeds
• 🤖 AI-powered summaries using Perplexity Sonar
• 💾 Local storage of all fetched articles and summaries
• 📊 Daily overview summary with bullet points
• 🔒 Budget management to protect API credits
• 📬 Automated Telegram delivery

Primary Constraint: Simple local execution with strict budget management to preserve the $5/month Perplexity API credit.

Prerequisites

• Python 3.10 or higher
• Perplexity API key (sign up at https://www.perplexity.ai/)
• Telegram Bot Token (create a bot using @BotFather on Telegram)
• Telegram Chat ID (use @userinfobot to get your chat ID, or use a channel ID)

Installation

1. Clone this repository:

   git clone https://github.com/soroush-thr/spectrum.git
   cd spectrum

2. Install dependencies:

   pip install -r requirements.txt

3. Set up environment variables:

   • Copy .env.example to .env:

     cp .env.example .env

   • Edit .env and fill in your API keys:

     PERPLEXITY_API_KEY=pplx-your-actual-key-here
     TELEGRAM_BOT_TOKEN=your-bot-token-here
     TELEGRAM_CHAT_ID=your-chat-id-here
     

Usage

Run the script:

python main.py

The script will:

1. Fetch articles from configured RSS feeds
2. Filter articles published in the last 48 hours (configurable)
3. Sort by date (newest first)
4. Process up to 5 articles (budget limit)
5. Summarize each article using Perplexity API
6. Post individual summaries to your Telegram channel/bot
7. Generate and send a final overview summary with bullet points
8. Save all fetched articles and summaries locally to storage/ directory

Output Format

Individual Messages:

Source: [Feed Name]

Headline: [Article Title]

Summary: [AI-generated summary]

Link: [Article URL]

Final Overview:

📊 Daily AI News Overview (YYYY-MM-DD)

• [Bullet point summary of all news]
• [Key developments and trends]
• [Consolidated insights]

Configuration

RSS Feeds

Edit config.py to modify the RSS feed sources. The default configuration includes:

• The Verge - AI
• TechCrunch - AI
• Ars Technica - AI
• MIT Technology Review - AI
• Wired - AI

Budget Management

The script includes built-in budget protection:

• MAX_ARTICLES: Maximum number of articles to process per run (default: 5)
• HOURS_LOOKBACK: Hours to look back for articles (default: 48)

These can be adjusted in config.py to match your budget and needs.

Local Storage

All fetched articles and generated summaries are automatically saved to the storage/ directory:

• Fetched Articles: storage/fetched_articles_YYYY-MM-DD_HH-MM-SS.json

  • Contains all articles found in RSS feeds (before filtering)
  • Includes title, link, description, publication date, and source feed

• Summaries: storage/summaries_YYYY-MM-DD_HH-MM-SS.json

  • Contains processed articles with AI-generated summaries
  • Includes all article metadata plus the summary and processing timestamp

This allows you to:

• Review what was fetched even if articles were too old to process
• Keep a local archive of all summaries
• Analyze trends over time
• Debug issues with specific articles

Budget Management Notes

• The script processes a maximum of 5 articles per run, even if more articles are found
• Articles are sorted by date (newest first) before applying the limit
• If an error occurs (e.g., API rate limit), the script will skip that article and continue with the next one
• Each run uses: (number of articles + 1) Perplexity API calls (one per article + one for final overview)
• Monitor your Perplexity API usage to stay within the $5/month credit

Troubleshooting

"PERPLEXITY_API_KEY not found in .env file"

• Make sure you've created a .env file (copy from .env.example)
• Verify the key is correctly formatted: PERPLEXITY_API_KEY=pplx-...

"TELEGRAM_BOT_TOKEN not found in .env file"

• Create a bot using @BotFather on Telegram
• Copy the token and add it to your .env file
• See telegram-bot-setup.md for detailed bot configuration instructions

"TELEGRAM_CHAT_ID not found in .env file"

• For personal chats: Use @userinfobot to get your chat ID
• For channels: Forward a message from your channel to @userinfobot or use the channel's username (e.g., @yourchannel)

"Error fetching feed"

• Check your internet connection
• Verify the RSS feed URL is accessible
• Some feeds may be temporarily unavailable
• The script will continue processing other feeds if one fails

"Perplexity API error: 429 Too Many Requests"

• You've hit the rate limit or exceeded your credit
• Wait before running the script again
• Check your Perplexity API usage dashboard
• Consider reducing MAX_ARTICLES in config.py

No articles found

• The RSS feeds may not have published new articles in the last 48 hours
• Check the feed URLs in config.py are correct and accessible
• Review the storage/fetched_articles_*.json files to see what was fetched
• Consider increasing HOURS_LOOKBACK in config.py if needed

File Structure

.
├── .env                  # Your local API keys (ignored by git)
├── .env.example          # Template for .env
├── .gitignore            # Excludes .env, __pycache__, storage/, etc.
├── config.py             # RSS feeds and constants
├── main.py               # Main application logic
├── requirements.txt      # Python dependencies
├── telegram-bot-setup.md # Telegram bot configuration guide
├── logo-prompt.md        # Logo generation prompt for bot profile picture
├── storage/              # Local storage directory (gitignored)
│   ├── fetched_articles_*.json
│   └── summaries_*.json
└── README.md             # This file

Development

This is a local MVP with no database or state management. The script relies on:

• Time-based filtering (last N hours) to avoid processing duplicate articles
• Local JSON files for storage and archival
• Simple error handling that continues processing on failures

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

See LICENSE file for details.