AgentPress - Multi-Agent AI Publishing House

AgentPress is a streamlined multi-agent AI publishing framework for creating high-quality stories. It leverages specialized AI agents that collaborate through detailed prompts to research, write, edit, and fact-check content. The system is specifically tailored for "Shan Hai Jing" stories but can be adapted to any narrative genre.

🌟 Key Features

• Modular Agent Architecture: Specialized AI agents in distinct roles:
  • 📚 Mythologist: Handles background research and mythological research
  • ✍️ Writer: Creates story content
  • 🔍 Fact Checker: Ensures narrative consistency and logic accuracy
  • 💬 Dialogue Specialist: Optimizes dialogue quality and character voices
  • 📝 Editor: Reviews overall story quality
  • 📋 Documentation Specialist: Maintains story consistency across chapters
• Multi-Chapter Support: Complete workflow supporting both single and multi-chapter stories with consistency tracking
• Iterative Review: Multi-round review and refinement until stories meet quality standards
• Prompt-Driven: All agent behaviors defined by detailed prompts for easy customization
• Chinese LLM Friendly: Compatible with Qwen and other OpenAI-compatible models
• Consistency Checking: Automatic consistency validation and updates across chapters

🚀 Quick Start

1. Clone the repository

git clone https://github.com/Ddhjx-code/AgentPress.git
cd AgentPress

2. Install dependencies

pip install -r requirements.txt
# Or manually:
pip install "autogen-agentchat>=0.7.5" "autogen-ext[openai]>=0.7.5" python-dotenv

3. Configure API key

Create .env file with:

QWEN_API_KEY=your_api_key_here

4. Run the example

python main.py

5. Check outputs in output/ directory

• output/novel_story_*.txt - Final story content
• output/novel_data_*.json - Complete structured data
• output/conversation_history_*.json - Full conversation records

📁 Project Structure

AgentPress/ ├── main.py # Entry point with async support ├── phases.py # Workflow orchestrator ├── src/ # Core modules directory │ ├── novel_phases_manager.py # Complete multi-phase writing implementation │ └── documentation_manager.py # Story consistency management ├── agents_manager.py # AI agent manager ├── conversation_manager.py # Communication and history manager ├── config.py # Configuration and settings ├── utils.py # Utility functions ├── prompts/ # Detailed agent prompts │ ├── mythologist.md # Background researcher system prompt │ ├── writer.md # Story writer system prompt │ ├── fact_checker.md # Fact checker system prompt │ ├── dialogue_specialist.md # Dialogue reviewer system prompt │ └── editor.md # Final editor system prompt ├── output/ # Generated content directory │ ├── novel_story_.txt # Story text files │ ├── novel_data_.json # Complete structured output │ └── conversation_history_*.json # Communication logs └── .env # API configuration

🎯 Use Cases

• Mythological Fiction: "Shan Hai Jing"-style stories
• Novel Writing: Multi-chapter story creation with consistency
• Content Creation: AI-assisted fiction writing
• Multi-Agent Collaboration: AI agent coordination research
• Prompt Engineering: High-quality prompt experimentation

🔄 Customizing Prompts

To adapt to different genres, modify prompt files in prompts/ directory:

• Change Setting: Modify mythology and background in mythologist.md
• Adjust Style: Update narrative requirements in writer.md
• Modify Genres: Change story types and tones across all prompts

📚 Architecture

This system uses a clean, modular architecture with clear separation of concerns:

• Orchestration layer: main.py and phases.py
• Business logic: src/novel_phases_manager.py
• Data management: src/documentation_manager.py and conversation_manager.py
• AI agents: agents_manager.py and prompts/

This structure allows for independent development and testing of components.

🤝 Contributing

Contributions are welcome! We especially welcome:

• New agent prompt designs
• Genre-specific configurations
• Multi-chapter consistency improvements
• Performance and reliability enhancements

📜 License

MIT License - see LICENSE for details