AI News Bot

An automated system that generates and distributes daily AI news digests using Anthropic's Claude API.

Features

• Multi-Provider LLM Support: Choose between Claude (Anthropic) or DeepSeek for news generation
• Real-Time News Fetching: Fetches actual news from RSS feeds for accurate, up-to-date content
• AI-Powered News Generation: Uses Claude Sonnet 4.5 or DeepSeek to generate comprehensive AI news digests
• Web Search Integration: Optional DuckDuckGo web search for additional news sources
• Beautiful Email Formatting: Automatically converts AI content to stunning HTML emails - no markdown, just clean professional design
• Customizable Prompts: 9 pre-built templates (comprehensive, research, business, technical, etc.) or create your own
• Multilingual Support: Generate news in 13+ languages including English, Chinese, Spanish, French, Japanese, and more
• Chinese News Sources: Built-in support for Chinese AI news sources (36Kr, JiQiZhiXin, etc.)
• Multiple Notification Channels: Supports email (via Resend.com) and webhook notifications
• Flexible Configuration: Easy-to-customize topics and notification settings via YAML config
• Automated Scheduling: GitHub Actions workflow for daily automated execution
• Robust Error Handling: Comprehensive logging and retry logic
• Email Client Compatible: Works perfectly in Gmail, Outlook, Apple Mail, and mobile email apps
• Modern Email Delivery: Uses Resend.com for reliable, developer-friendly email delivery

🚀 Deployment Options

Choose your deployment method:

Method	Configuration	When to Use
Local Development	.env file	Testing locally on your computer
GitHub Actions	Repository Secrets	Automated daily runs (recommended)

💡 Tip: Start with local development to test, then deploy to GitHub Actions for automation.

Quick Start (Local Development)

1. Clone the Repository

git clone <your-repo-url>
cd ai-news-bot

2. Install Dependencies

pip install -r requirements.txt

3. Configure Settings (For Local Development)

For local development, copy the example file and fill in your credentials:

cp .env.example .env

Edit .env with your actual values:

# LLM Provider Configuration
LLM_PROVIDER=claude  # Options: 'claude' or 'deepseek'

# API Keys (provide the one you're using)
ANTHROPIC_API_KEY=your_api_key_here      # For Claude
DEEPSEEK_API_KEY=your_deepseek_api_key   # For DeepSeek

# Optional: Email Configuration with Resend.com
RESEND_API_KEY=re_your_api_key_here
EMAIL_FROM=your_email@yourdomain.com
EMAIL_TO=recipient@example.com

# Optional: Webhook Configuration
WEBHOOK_URL=https://your-webhook-url.com/endpoint

# Notification Methods (comma-separated)
NOTIFICATION_METHODS=email,webhook

# Language Settings (optional, defaults to 'en')
AI_RESPONSE_LANGUAGE=zh

# Web Search (optional, defaults to false)
ENABLE_WEB_SEARCH=false

Note: The .env file is only for local development. For GitHub Actions automation, you'll configure these as GitHub Secrets (see GitHub Actions Setup below).

4. Customize News Prompt (Optional)

The bot uses an optimized, concise prompt (15 lines vs 50+ in typical systems) that generates high-quality news digests.

Default Prompt (in config.yaml):

Summarize 10 recent AI news items (5 international + 5 domestic) covering: {topics}

Format:
International News:
1. [Headline]
[2-3 sentence summary]
Source: [Name]

Domestic News:
1. [Headline]
...

Rules: Recent news, no markdown, clear language

Why it's concise:

• ✅ Faster processing
• ✅ Lower cost
• ✅ Easier to maintain
• ✅ No redundancy

Multi-Language Support:

Prompts are in English (best for Claude), but output can be in 13+ languages:

# In .env file
AI_RESPONSE_LANGUAGE=zh  # Chinese output
AI_RESPONSE_LANGUAGE=es  # Spanish output
AI_RESPONSE_LANGUAGE=ja  # Japanese output
# Supports: en, zh, es, fr, ja, de, ko, pt, ru, ar, hi, it, nl

Pre-built Templates (config.examples.yaml):

1. Comprehensive (default) - Balanced coverage
2. Research - Academic focus
3. Business - Industry & funding
4. Technical - Engineering depth
5. Startup - Early-stage companies
6. Policy - Regulations
7. Weekly - Top stories
8. Concise - Ultra-brief
9. Chinese - 中文示例

📖 Full Guide: See config.examples.yaml for customization and multi-language details.

5. Run Locally

python main.py

Configuration

Configuration Variables

The bot requires the following configuration. How you set them depends on your deployment:

• Local Development: Use .env file (see Quick Start)
• GitHub Actions: Use GitHub Repository Secrets (see GitHub Actions Setup)

Variable	Required	Description
LLM_PROVIDER	Optional	LLM provider to use: claude or deepseek (default: claude)
ANTHROPIC_API_KEY	If using Claude	Your Anthropic API key
DEEPSEEK_API_KEY	If using DeepSeek	Your DeepSeek API key
NOTIFICATION_METHODS	✅ Required	Comma-separated list: email, webhook, or email,webhook
AI_RESPONSE_LANGUAGE	Optional	Language code for AI responses (default: en). Supports: zh, es, fr, ja, de, ko, pt, ru, ar, hi, it, nl
ENABLE_WEB_SEARCH	Optional	Enable web search for news (default: false)
RESEND_API_KEY	If using email	Your Resend.com API key
EMAIL_FROM	If using email	Sender email address (must be verified in Resend)
EMAIL_TO	If using email	Recipient email address
WEBHOOK_URL	If using webhook	Webhook endpoint URL

Configuration File (config.yaml)

The config.yaml file allows you to customize the news digest behavior:

LLM Configuration:

• Provider: Choose between claude or deepseek
• Model: Optionally specify a specific model version

News Configuration:

• use_real_sources: Enable fetching news from RSS feeds (recommended, default: true)
• enable_web_search: Enable DuckDuckGo web search (default: false)
• include_chinese_sources: Include Chinese news sources (default: true)
• max_items_per_source: Maximum news items per source (default: 10)
• Topics: Focus areas for news selection (optional, guides the AI)
• Prompt Template: The instruction template for the LLM
  • Default: Comprehensive 15-20 item digest with category headers
  • Fully customizable with your own prompts
  • See config.examples.yaml for 9 pre-built templates

Logging Settings: Control log verbosity and format

Example Structure:

llm:
  provider: claude  # or 'deepseek'
  # model: claude-sonnet-4-5-20250929  # optional

news:
  use_real_sources: true
  enable_web_search: false
  include_chinese_sources: true
  max_items_per_source: 10

  topics:
    - "Large Language Models (LLM)"
    - "AI Agents and Autonomous Systems"
    - "Product launches"

  prompt_template: |
    Your custom prompt...
    Focus: {topics}

logging:
  level: INFO
  format: "%(asctime)s - %(levelname)s - %(message)s"

LLM Provider Configuration

The bot supports multiple LLM providers. Configure in config.yaml or via environment variables:

Claude (Anthropic) - Default

llm:
  provider: claude
  model: claude-sonnet-4-5-20250929  # optional, uses default if not set

Available Claude Models:

• claude-sonnet-4-5-20250929 - Latest Sonnet (default) - Best for most tasks
• claude-3-5-sonnet-20241022 - Previous Sonnet version

Claude Pricing (per million tokens):

• Claude Sonnet 4.5: $3 input / $15 output

DeepSeek - Cost-Effective Alternative

llm:
  provider: deepseek
  model: deepseek-chat  # optional, uses default if not set

Available DeepSeek Models:

• deepseek-chat - General chat model (default)
• deepseek-reasoner - Enhanced reasoning model

DeepSeek Pricing:

• Much lower cost than Claude
• Better Chinese language support
• Good quality for news summarization

Choosing a Provider

Provider	Pros	Cons	Best For
Claude	Excellent quality, reliable	Higher cost	Production, high-quality output
DeepSeek	Low cost, good Chinese	Slightly lower quality	Budget-conscious, Chinese content

Language Configuration

How It Works:

• Prompts are always in English (best for Claude understanding)
• Output can be in 13+ languages (automatic translation)
• Set AI_RESPONSE_LANGUAGE in .env or GitHub Secrets

Supported Languages:

en (English) • zh (中文) • es (Español) • fr (Français) • ja (日本語) • de (Deutsch) • ko (한국어) • pt (Português) • ru (Русский) • ar (العربية) • hi (हिन्दी) • it (Italiano) • nl (Nederlands)

Usage:

# .env file
AI_RESPONSE_LANGUAGE=zh  # Full Chinese output

# GitHub Secret
# Add: AI_RESPONSE_LANGUAGE = zh

Example Output (Chinese):

国际新闻：

1. OpenAI发布GPT-5增强推理能力
OpenAI发布了GPT-5...
来源：OpenAI官方博客

The system automatically adds: "IMPORTANT: Please respond entirely in Chinese (中文)" to the prompt.

GitHub Actions Setup

The project includes a GitHub Actions workflow that runs daily at midnight UTC (00:00).

Important: GitHub Actions uses Repository Secrets for configuration (NOT environment variables). All settings must be added as secrets.

Step 1: Add GitHub Repository Secrets

Navigate to your GitHub repository:

Repository → Settings → Secrets and variables → Actions → Repository secrets → New repository secret

Add the following secrets one by one:

✅ Required Secrets

Secret Name	Example Value	Description
LLM_PROVIDER	claude or deepseek	LLM provider to use (default: claude)
ANTHROPIC_API_KEY	sk-ant-api03-xxx...	Your Anthropic API key (if using Claude)
DEEPSEEK_API_KEY	sk-xxx...	Your DeepSeek API key (if using DeepSeek)
NOTIFICATION_METHODS	email,webhook	Notification channels (comma-separated)

📧 Email Secrets (if using email notifications)

Secret Name	Example Value	Description
RESEND_API_KEY	re_123abc...	Your Resend.com API key
EMAIL_FROM	news@yourdomain.com	Sender email (must be verified in Resend)
EMAIL_TO	you@example.com	Recipient email address

🔗 Webhook Secrets (if using webhook notifications)

Secret Name	Example Value	Description
WEBHOOK_URL	https://hooks.slack.com/...	Your webhook endpoint URL

🌍 Optional Secrets

Secret Name	Example Value	Description
AI_RESPONSE_LANGUAGE	zh or es or ja	Language code (defaults to en if not set)
ENABLE_WEB_SEARCH	true or false	Enable web search for news (defaults to false)

Step 2: Enable GitHub Actions

Ensure GitHub Actions are enabled in your repository settings:

Repository → Settings → Actions → General → Allow all actions and reusable workflows

Step 3: Manual Trigger (Test Your Setup)

Once secrets are configured, test your setup:

Repository → Actions tab → Daily AI News Digest → Run workflow button

This will run the workflow immediately so you can verify everything is working.

Step 4: Customize Schedule (Optional)

The workflow runs daily at midnight UTC by default. To change the schedule, edit .github/workflows/daily-news.yml:

schedule:
  - cron: '0 0 * * *'  # Midnight UTC daily (current)
  - cron: '0 9 * * *'  # 9:00 AM UTC daily
  - cron: '0 */6 * * *'  # Every 6 hours

Use crontab.guru to create custom schedules.

Project Structure

ai-news-bot/
├── .github/
│   └── workflows/
│       └── daily-news.yml           # GitHub Actions workflow
├── src/
│   ├── __init__.py
│   ├── config.py                    # Configuration management
│   ├── logger.py                    # Logging utilities
│   ├── news_generator.py            # News generation orchestration
│   ├── news_fetcher.py              # RSS feed news fetching
│   ├── web_search.py                # DuckDuckGo web search integration
│   ├── llm_providers/
│   │   ├── __init__.py
│   │   ├── base_provider.py         # Base LLM provider interface
│   │   ├── claude_provider.py       # Anthropic Claude provider
│   │   └── deepseek_provider.py     # DeepSeek provider
│   └── notifiers/
│       ├── __init__.py
│       ├── email_notifier.py        # Email notification
│       └── webhook_notifier.py      # Webhook notification
├── main.py                          # Main application entry point
├── config.yaml                      # Active configuration file
├── requirements.txt                 # Python dependencies
├── .env.example                     # Example environment variables
├── .gitignore
├── README.md
└── README.zh.md                     # Chinese documentation

Usage Examples

Email Only

NOTIFICATION_METHODS=email

Webhook Only

NOTIFICATION_METHODS=webhook

Both Email and Webhook

NOTIFICATION_METHODS=email,webhook

Email Format

Beautiful, Email-Friendly Design

The bot generates email-optimized content that looks stunning across all email clients:

Features:

• ✅ No markdown formatting (clean, professional appearance)
• ✅ Automatic HTML conversion with beautiful styling
• ✅ Numbered news cards with visual badges
• ✅ Color-coded sections and headers
• ✅ Mobile-responsive layout
• ✅ Works in Gmail, Outlook, Apple Mail, and all mobile apps

What recipients see:

• Clean white container with professional styling
• Blue section headers with subtle borders
• Numbered news items in styled cards
• Italicized source citations
• Comfortable reading experience on any device

Preview your emails: Run the bot locally and check the generated HTML email content.

Email Setup Guide with Resend.com

Setting Up Resend

1. Sign up for Resend

   • Go to resend.com and create an account
   • Resend offers a generous free tier (100 emails/day, 3,000 emails/month)

2. Get Your API Key

   • Navigate to API Keys in your Resend dashboard
   • Create a new API key
   • Copy the API key (starts with re_) and set it as RESEND_API_KEY

3. Verify Your Domain (Recommended for production)

   • Go to Domains in your Resend dashboard
   • Add and verify your domain by adding DNS records
   • Once verified, you can send from any address at your domain

4. For Testing (No domain verification needed)

   • You can use onboarding@resend.dev as the EMAIL_FROM address
   • This is only for testing and has sending limits
   • For production use, verify your own domain

Why Resend?

• Simple API: Easy-to-use REST API, much simpler than SMTP
• Better Deliverability: Higher inbox placement rates
• No SMTP Configuration: No need to manage SMTP credentials
• Modern: Built for developers with excellent documentation
• Analytics: Track email delivery and engagement

Webhook Integration

The webhook sends a JSON payload:

{
  "title": "AI News Digest - 2025-10-25",
  "content": "... news digest content ...",
  "timestamp": "2025-10-25T09:00:00",
  "source": "AI News Bot"
}

Compatible with:

• Slack (use Incoming Webhooks)
• Discord (use Webhook URLs)
• Microsoft Teams
• Custom webhook endpoints

Error Handling

• Automatic Retries: The news generator retries up to 3 times on failure
• Graceful Degradation: If one notification method fails, others still execute
• Comprehensive Logging: All operations are logged with timestamps and context
• GitHub Actions Artifacts: Error logs are uploaded for debugging

Troubleshooting

"Config file not found" Error

Ensure config.yaml exists in the project root.

Email Not Sending

• Verify RESEND_API_KEY is correct and active
• Ensure EMAIL_FROM is verified in Resend (or use onboarding@resend.dev for testing)
• Check Resend dashboard for delivery logs and errors
• Verify you haven't exceeded Resend's sending limits

Webhook Failing

• Verify webhook URL is accessible
• Check webhook endpoint accepts JSON POST requests
• Review webhook service logs

API Errors

• Verify ANTHROPIC_API_KEY is valid
• Check API quota/rate limits
• Review Anthropic API status

Development

Running Tests (when available)

pytest

Local Development

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Run the bot
python main.py

License

MIT License - See LICENSE file for details

Support

For issues and feature requests, please use the GitHub issue tracker.

Credits

Powered by:

• Anthropic Claude
• DeepSeek