Google Gemini Integration for UglyFeed

Overview

Google Gemini AI support has been successfully added to UglyFeed, providing access to Google's advanced language models for content processing and rewriting.

Features Added

1. Gemini API Client

• Full Gemini API integration with proper error handling and retry logic
• Content truncation to handle token limits
• Rate limiting support with automatic retry
• Proper response parsing for Gemini's unique response format

2. Configuration Support

• GUI Integration: Added Gemini as an option in the Streamlit interface
• Config file support: Example configuration in config.yaml
• Environment variable support: Can be configured via environment variables
• CLI support: Command-line interface supports Gemini parameters

3. Supported Models (Free Tier Available!)

Free Tier Models (No cost!)

• gemini-2.0-flash-exp ⭐ RECOMMENDED - Latest experimental model with enhanced capabilities
• gemini-2.0-flash - Latest stable model with excellent performance
• gemini-1.5-flash - Proven fast model with 1M token context window
• gemini-1.5-flash-8b - Smallest model for simple tasks
• gemma-2-2b-it - Open source instruction-tuned model
• gemma-2-9b-it - Larger open source model

Paid Tier Models

• gemini-1.5-pro - High intelligence with 2M token context
• gemini-2.0-flash-thinking-exp - Advanced reasoning capabilities

💡 Tip: Start with gemini-2.0-flash-exp - it's free and offers the latest capabilities!

Configuration

Via config.yaml

api_config:
  selected_api: "Gemini"
  gemini_api_url: "https://generativelanguage.googleapis.com/v1beta/models/{model}:generateContent"
  gemini_api_key: "your_gemini_api_key"
  gemini_model: "gemini-2.0-flash-exp"  # Latest experimental model

Via Environment Variables

export API_TYPE="gemini"
export API_KEY="your_gemini_api_key"
export API_MODEL="gemini-2.0-flash-exp"
export API_URL="https://generativelanguage.googleapis.com/v1beta/models/{model}:generateContent"

Via Command Line

python llm_processor.py --api gemini \
                       --api_key "your_gemini_api_key" \
                       --model "gemini-2.0-flash" \
                       --api_url "https://generativelanguage.googleapis.com/v1beta/models/{model}:generateContent"

Getting Started

1. Get Gemini API Key

1. Go to Google AI Studio
2. Create a new API key
3. Copy the key for use in UglyFeed

2. Configure UglyFeed

Choose one of the configuration methods above and add your API key.

3. Run UglyFeed

• GUI: Select "Gemini" from the API dropdown in the Streamlit interface
• CLI: Use the command-line parameters shown above
• Automated: Use the provided GitHub Action workflow

GitHub Actions Integration

A ready-to-use GitHub Actions workflow is available at: docs/UglyFeed-GitHub-Action-Gemini-gemini-2.0-flash.yml

This workflow uses the free tier gemini-2.0-flash model.

To use it:

1. Add your Gemini API key as a GitHub secret named GEMINI_API_KEY
2. Copy the workflow to .github/workflows/ in your repository
3. The workflow will run every 4 hours automatically using the free tier

Technical Details

API Implementation

• Base URL: https://generativelanguage.googleapis.com/v1beta/models/{model}:generateContent
• Authentication: API key passed as query parameter
• Request Format: JSON with contents array
• Response Format: JSON with candidates array containing generated content

Error Handling

• Rate limiting: Automatic retry with exponential backoff
• API errors: Comprehensive error logging with response details
• Network issues: Robust retry mechanism with session management
• Content validation: Proper response parsing with fallback handling

Token Management

• Content truncation: Automatically truncates content to fit within token limits
• Token estimation: Rough estimation for content length management
• Configurable limits: Adjustable maximum token limits via configuration

Troubleshooting

Common Issues

1. "API key not provided": Ensure your API key is properly set in configuration
2. "Model not found": Verify the model name is correct (e.g., "gemini-1.5-flash")
3. "Rate limit exceeded": The system will automatically retry, or wait a few minutes

Validation

The system includes comprehensive validation:

• API key presence check
• Model name validation
• URL format verification
• Response structure validation

Migration from Other APIs

Switching from other APIs to Gemini is seamless:

1. Change selected_api to "Gemini" in config.yaml
2. Add Gemini-specific configuration parameters
3. Remove or comment out previous API configuration
4. Restart UglyFeed

Performance Notes

• gemini-2.0-flash: ⭐ FREE & RECOMMENDED - Best balance of speed and quality
• gemini-2.0-flash-lite: FREE - Fastest response times, good for high-volume processing
• gemini-1.5-flash: FREE - Proven performance with 1M token context
• gemini-1.5-flash-8b: FREE - Most efficient for simple tasks
• gemma-3/gemma-3n: COMPLETELY FREE - Open source models
• Rate limits: Gemini free tier has generous limits with automatic handling

Cost Benefits

🎉 All recommended models are completely FREE!

• No credit card required for free tier
• Generous rate limits for development and small-scale production
• Perfect for RSS feed processing and content rewriting
• Data used to improve Google products (standard for free tier)

Security

• API keys are never logged or exposed in output
• HTTPS-only communication with Google's servers
• Proper error handling prevents sensitive data leakage