Telegram Easy Summary

A tool that creates AI-powered summaries of Telegram chat messages using OpenRouter AI models.

Features

• Fetch messages from Telegram channels/chats
• Focus on specific users while maintaining conversation context
• Generate summaries using any model from OpenRouter (default: GPT-3.5 Turbo)
• Analyze messages by participant with individual summaries
• Fetch only unread messages with a single flag
• Auto-posting: Automatically post analysis of unread messages to a Telegram channel every N hours

Setup

1. Clone this repository

2. Install dependencies: pip install -r requirements.txt

3. Configure .env file:

   TELEGRAM_API_ID=your_api_id
   TELEGRAM_API_HASH=your_api_hash
   TELEGRAM_STRING_SESSION=your_session_string
   DEFAULT_TELEGRAM_CHANNEL_ID=your_default_channel_id
   OPENROUTER_API_KEY=your_openrouter_api_key
   AUTO_POSTING_CHANNEL_ID=your_target_channel_id  # Optional: for auto-posting
   

4. For auto-posting: Configure config.yaml:

   auto_posting:
     enabled: true
     interval_hours: 24  # Post every 24 hours
     target_channel_id: -1001234567890  # Channel to post to
     format: "markdown"  # or "text" or "json"
     unread_only: true

Usage

# Basic usage with default settings
python main.py

# Custom examples
python main.py -n 200                              # Fetch 200 messages
python main.py -c @channel_name                    # Specify channel
python main.py -u @username1 @username2            # Focus on users
python main.py -o summary.txt                      # Save to file
python main.py --unread                            # Only unread messages
python main.py --model anthropic/claude-3-5-sonnet  # Use Claude model from OpenRouter
python main.py --model openai/gpt-4o              # Use GPT-4o model from OpenRouter
python main.py --auto-post                        # Start auto-posting scheduler

Command Options

• -c, --chat_id: Chat ID or username (default: from config)
• -n, --limit: Max messages to fetch (default: 100)
• -u, --users: Users to focus on
• -o, --output: Output file path
• -m, --model: OpenRouter model for summarization
• --unread: Fetch only unread messages
• --auto-post: Start auto-posting scheduler (runs continuously)

Customizing Prompts

The tool uses separate markdown files for prompts located in data/prompts/:

• example_prompt.md-example: Main prompt for generating the overall conversation summary

To customize prompts:

1. Save as the relevant .md file in data/prompts/
2. Your changes will be automatically used when running the tool
3. Use template variables like {participants} and {messages} in your prompts
4. In your config.yaml file, set the prompt_template_name under default_prompt to match your preferred prompt file name (without the .md extension). For example, if you want to use summary_prompt.md, set it to 'summary_prompt'

Configuration

Edit config.yaml to change default settings like message limits, default channel, and model selection.

OpenRouter Models

The configuration includes these models that you can use:

1. openai/o4-mini - OpenAI's O4 Mini model
2. anthropic/claude-3.7-sonnet - Anthropic's Claude 3.7 Sonnet model
3. google/gemini-2.5-flash-preview-05-20 - Google's Gemini 2.5 Flash Preview

OpenRouter Credits

To use this tool, you need OpenRouter credits. To replenish your credits:

1. Visit https://openrouter.ai/settings/credits
2. Add more credits to your account using available payment methods

License

MIT