🧠 ArgumentRx - Retrieval-Augmented Argumentation System

ArgumentRx is an AI-powered system that generates fact-based, pro/con arguments on controversial or complex topics by retrieving real-time evidence from free internet sources. It combines state-of-the-art retrieval techniques with large language models (LLMs) to mimic human-style critical thinking.

🌟 Features

• Real-time Source Retrieval: Automatically gathers evidence from news APIs, Reddit, and crypto research sources
• Structured Argument Generation: Creates logically sound, stance-based arguments using retrieved content
• Built-in Evaluation: Scores argument strength using rule-based logic and optional LLM feedback
• Source Transparency: Shows all sources and evidence clearly for responsible AI use
• Interactive Web Interface: Streamlit-based UI for easy interaction
• Export Capabilities: Save results as JSON or generate transparency reports

🚀 Quick Start

1. Installation

git clone https://github.com/ysimokat/Crypto_RAG.git
cd Crypto_RAG
pip install -r requirements.txt

2. Configuration

Copy the example environment file and add your API keys:

cp .env.example .env
# Edit .env with your API keys

Required API keys (optional but recommended):

• OPENAI_API_KEY: For GPT-based argument generation
• REDDIT_CLIENT_ID & REDDIT_CLIENT_SECRET: For Reddit content retrieval
• NEWS_API_KEY: For news article retrieval

3. Usage Options

Web Interface (Recommended)

streamlit run streamlit_app.py

Command Line Interface

python src/argumentrx.py "Bitcoin environmental impact"
python src/argumentrx.py "DeFi regulation" --save results.json --transparency

Python API

from src.argumentrx import ArgumentRx

argumentrx = ArgumentRx()
results = argumentrx.generate_arguments("Proof of Work vs Proof of Stake")
print(results)

📁 Project Structure

Crypto_RAG/
├── src/
│   ├── argumentrx.py          # Main ArgumentRx class
│   ├── retrieval.py           # Data retrieval from various sources
│   ├── embeddings.py          # Text embedding and similarity search
│   ├── argument_generator.py  # LLM-based argument generation
│   └── evaluator.py          # Argument quality evaluation
├── streamlit_app.py           # Web interface
├── example_usage.py           # Usage examples
├── config.py                  # Configuration settings
├── requirements.txt           # Python dependencies
└── README.md                  # This file

🛠️ Core Components

1. Data Retrieval (retrieval.py)

• NewsRetriever: Fetches articles from news APIs
• RedditRetriever: Searches relevant subreddits
• CryptoBlogRetriever: Gathers content from crypto news sites
• UnifiedRetriever: Coordinates all retrieval sources

2. Embedding Management (embeddings.py)

• Uses sentence transformers for text embedding
• FAISS-based similarity search for relevant source selection
• Persistent index storage and loading

3. Argument Generation (argument_generator.py)

• ArgumentGenerator: OpenAI GPT-based generation
• LocalLLMGenerator: Local model support (HuggingFace)
• Structured pro/con argument creation with citations

4. Evaluation System (evaluator.py)

• Rule-based scoring (clarity, logic, evidence, persuasiveness)
• Optional LLM-based evaluation for enhanced feedback
• Combined scoring methodology

📊 Evaluation Metrics

Arguments are scored on four dimensions:

• Clarity (0.0-1.0): How clear and understandable the argument is
• Logic (0.0-1.0): Logical structure and reasoning flow
• Evidence (0.0-1.0): Use of citations and factual support
• Persuasiveness (0.0-1.0): Convincing power and rhetorical effectiveness

🎯 Use Cases

• Crypto Research: Analyze pros/cons of tokens, protocols, or regulations
• Debate Preparation: Train students to see multiple perspectives
• Policy Analysis: Summarize opposing viewpoints on laws and governance
• Research Assistant: Demonstrate transparent reasoning systems

📋 Example Output

{
  "topic": "Bitcoin environmental impact",
  "arguments": {
    "pro": {
      "content": "Bitcoin mining drives renewable energy adoption...",
      "evaluation": {
        "overall_score": 0.75,
        "clarity": 0.8,
        "logic": 0.7,
        "evidence": 0.8,
        "persuasiveness": 0.7
      },
      "citations": ["[1] Energy consumption study - coindesk.com"]
    },
    "con": {
      "content": "Bitcoin's energy consumption exceeds many countries...",
      "evaluation": {
        "overall_score": 0.82,
        "clarity": 0.85,
        "logic": 0.8,
        "evidence": 0.85,
        "persuasiveness": 0.78
      }
    }
  },
  "metadata": {
    "total_sources_retrieved": 25,
    "sources_used": 10,
    "model_used": "OpenAI"
  }
}

🔧 Configuration

Key settings in config.py:

• EMBEDDING_MODEL: Sentence transformer model for embeddings
• LLM_MODEL: Language model for argument generation
• RETRIEVAL_TOP_K: Number of sources to use for arguments
• MAX_SOURCES_PER_ARGUMENT: Maximum citations per argument
• ARGUMENT_MAX_LENGTH: Maximum argument length in words

🧪 Examples

Run the example script to see ArgumentRx in action:

python example_usage.py

This demonstrates:

• Basic argument generation
• File output and transparency reports
• Multiple topic analysis
• Evaluation breakdowns

⚡ GPU Support

The system automatically detects and uses GPU when available:

• Embedding models run on GPU for faster processing
• Local LLM generation utilizes GPU acceleration
• For optimal performance, install: pip install faiss-gpu

🚨 Limitations

• Requires internet connection for source retrieval
• API rate limits may affect performance
• Local LLM quality depends on model choice
• Source availability varies by topic

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

📄 License

This project is licensed under the MIT License.

🙏 Acknowledgments

• Built with HuggingFace Transformers, OpenAI API, and FAISS
• Inspired by retrieval-augmented generation research
• Uses public APIs for news, Reddit, and crypto content

---

Note: This system is designed for educational and research purposes. Always verify information from multiple sources and consider the limitations of AI-generated content.