ai-coscientist

AI-CoScientist

A multi-agent AI framework for collaborative scientific research, implementing the "Towards an AI Co-Scientist" methodology with tournament-based hypothesis evolution, peer review systems, and intelligent agent orchestration.

Features

🧠 Multi-Agent Architecture: Specialized agents for hypothesis generation, peer review, ranking, evolution, and meta-analysis
🏆 Tournament-Based Selection: Elo rating system for hypothesis ranking through pairwise comparisons
📊 Comprehensive Review System: Scientific soundness, novelty, testability, and impact assessment
🔄 Iterative Refinement: Meta-review guided evolution with strategic hypothesis improvement
🎯 Diversity Control: Proximity analysis to maintain hypothesis diversity and reduce redundancy
📈 Execution Metrics: Detailed performance tracking and agent timing analytics
💾 State Persistence: Save and resume research workflows with agent state management
🛡️ Robust Error Handling: Graceful fallbacks and recovery mechanisms for production reliability

Installation

You can install the package using pip:

pip install -e .

Or install dependencies directly:

pip install swarms loguru python-dotenv

Quick Start

from ai_coscientist import AIScientistFramework

# Initialize the AI Co-scientist Framework
ai_coscientist = AIScientistFramework(
    model_name="gpt-4",
    max_iterations=3,
    hypotheses_per_generation=10,
    tournament_size=8,
    evolution_top_k=3,
    verbose=True
)

# Define your research goal
research_goal = "Develop novel approaches for improving reasoning capabilities in large language models"

# Run the research workflow
results = ai_coscientist.run_research_workflow(research_goal)

# Access the results
print(f"Generated {len(results['top_ranked_hypotheses'])} top hypotheses")
for i, hypothesis in enumerate(results['top_ranked_hypotheses'], 1):
    print(f"{i}. {hypothesis['text']}")
    print(f"   Elo Rating: {hypothesis['elo_rating']}")
    print(f"   Win Rate: {hypothesis['win_rate']}%")

Architecture

The AI-CoScientist framework consists of 8 specialized agents:

• Generation Agent: Creates novel research hypotheses
• Reflection Agent: Peer review and scientific critique
• Ranking Agent: Hypothesis ranking and selection
• Evolution Agent: Hypothesis refinement and improvement
• Meta-Review Agent: Cross-hypothesis insight synthesis
• Proximity Agent: Similarity analysis and diversity control
• Tournament Agent: Pairwise hypothesis comparison
• Supervisor Agent: Workflow orchestration and planning

Advanced Usage

Custom Configuration

ai_coscientist = AIScientistFramework(
    model_name="claude-3-sonnet",
    max_iterations=5,
    base_path="./custom_states",
    verbose=True,
    tournament_size=12,
    hypotheses_per_generation=15,
    evolution_top_k=5,
)

State Management

# Save agent states
ai_coscientist.save_state()

# Load previous states
ai_coscientist.load_state()

Results Analysis

results = ai_coscientist.run_research_workflow(research_goal)

# Execution metrics
metrics = results['execution_metrics']
print(f"Total time: {results['total_workflow_time']:.2f}s")
print(f"Hypotheses generated: {metrics['hypothesis_count']}")
print(f"Reviews completed: {metrics['reviews_count']}")
print(f"Tournament rounds: {metrics['tournaments_count']}")

# Meta-review insights
insights = results['meta_review_insights']
print("Strategic recommendations:", insights.get('strategic_recommendations'))

Code Quality 🧹

• make style to format the code
• make check_code_quality to check code quality (PEP8 basically)
• black .
• ruff . --fix

Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

Documentation

For detailed documentation, see DOCS.md.

Citation

If you use AI-CoScientist in your research, please cite:

@software{ai_coscientist,
  title={AI-CoScientist: A Multi-Agent Framework for Collaborative Scientific Research},
  author={The Swarm Corporation},
  year={2024},
  url={https://github.com/The-Swarm-Corporation/AI-CoScientist}
}

License

MIT License - see LICENSE file for details.

Support

• 📧 Email: kye@swarms.world
• 💬 Discord: Join our community
• 🐦 Twitter: @kyegomezb