{
  "id": "57bef5a4-1e47-4957-b467-fe91c32ab920",
  "name": "AI-CoScientist",
  "title": "Agent",
  "description": "An simple, reliable, and minimal implementation of the AI CoScientist Paper from Google \"Towards an AI co-scientist\" with Swarms Framework\n\n![AI-CoScientist](https://storage.googleapis.com/gweb-research2023-media/images/AICoScientist-1-Components.width-1250.png)\n\n# AI-CoScientist\n\n[![Join our Discord](https://img.shields.io/badge/Discord-Join%20our%20server-5865F2?style=for-the-badge&logo=discord&logoColor=white)](https://discord.gg/swarms-999382051935506503) [![Subscribe on YouTube](https://img.shields.io/badge/YouTube-Subscribe-red?style=for-the-badge&logo=youtube&logoColor=white)](https://www.youtube.com/@kyegomez3242) [![Connect on LinkedIn](https://img.shields.io/badge/LinkedIn-Connect-blue?style=for-the-badge&logo=linkedin&logoColor=white)](https://www.linkedin.com/in/kye-g-38759a207/) [![Follow on X.com](https://img.shields.io/badge/X.com-Follow-1DA1F2?style=for-the-badge&logo=x&logoColor=white)](https://x.com/kyegomezb)\n\nA 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.\n\n## Features\n\n🧠 **Multi-Agent Architecture**: Specialized agents for hypothesis generation, peer review, ranking, evolution, and meta-analysis  \n🏆 **Tournament-Based Selection**: Elo rating system for hypothesis ranking through pairwise comparisons  \n📊 **Comprehensive Review System**: Scientific soundness, novelty, testability, and impact assessment  \n🔄 **Iterative Refinement**: Meta-review guided evolution with strategic hypothesis improvement  \n🎯 **Diversity Control**: Proximity analysis to maintain hypothesis diversity and reduce redundancy  \n📈 **Execution Metrics**: Detailed performance tracking and agent timing analytics  \n💾 **State Persistence**: Save and resume research workflows with agent state management  \n🛡️ **Robust Error Handling**: Graceful fallbacks and recovery mechanisms for production reliability\n\n## Installation\n\nYou can install the package using pip:\n\n\n```bash\npip3 install -U ai-coscientist\n```\n\n\n---\n\n\n## Quick Start\n\n```python\nfrom ai_coscientist import AIScientistFramework\n\n# Initialize the AI Co-scientist Framework\nai_coscientist = AIScientistFramework(\n    model_name=\"gpt-4o-mini\",\n    max_iterations=3,\n    hypotheses_per_generation=10,\n    tournament_size=8,\n    evolution_top_k=3,\n    verbose=True\n)\n\n# Define your research goal\nresearch_goal = \"Develop novel approaches for improving reasoning capabilities in large language models\"\n\n# Run the research workflow\nresults = ai_coscientist.run_research_workflow(research_goal)\n\n# Access the results\nprint(f\"Generated {len(results['top_ranked_hypotheses'])} top hypotheses\")\nfor i, hypothesis in enumerate(results['top_ranked_hypotheses'], 1):\n    print(f\"{i}. {hypothesis['text']}\")\n    print(f\"   Elo Rating: {hypothesis['elo_rating']}\")\n    print(f\"   Win Rate: {hypothesis['win_rate']}%\")\n```\n\n## Architecture\n\nThe AI-CoScientist framework consists of 8 specialized agents:\n\n- **Generation Agent**: Creates novel research hypotheses\n- **Reflection Agent**: Peer review and scientific critique\n- **Ranking Agent**: Hypothesis ranking and selection\n- **Evolution Agent**: Hypothesis refinement and improvement\n- **Meta-Review Agent**: Cross-hypothesis insight synthesis\n- **Proximity Agent**: Similarity analysis and diversity control\n- **Tournament Agent**: Pairwise hypothesis comparison\n- **Supervisor Agent**: Workflow orchestration and planning\n\n## Advanced Usage\n\n### Custom Configuration\n\n```python\nai_coscientist = AIScientistFramework(\n    model_name=\"claude-3-sonnet\",\n    max_iterations=5,\n    base_path=\"./custom_states\",\n    verbose=True,\n    tournament_size=12,\n    hypotheses_per_generation=15,\n    evolution_top_k=5,\n)\n```\n\n### State Management\n\n```python\n# Save agent states\nai_coscientist.save_state()\n\n# Load previous states\nai_coscientist.load_state()\n```\n\n### Results Analysis\n\n```python\nresults = ai_coscientist.run_research_workflow(research_goal)\n\n# Execution metrics\nmetrics = results['execution_metrics']\nprint(f\"Total time: {results['total_workflow_time']:.2f}s\")\nprint(f\"Hypotheses generated: {metrics['hypothesis_count']}\")\nprint(f\"Reviews completed: {metrics['reviews_count']}\")\nprint(f\"Tournament rounds: {metrics['tournaments_count']}\")\n\n# Meta-review insights\ninsights = results['meta_review_insights']\nprint(\"Strategic recommendations:\", insights.get('strategic_recommendations'))\n```\n\n## Code Quality 🧹\n\n- `make style` to format the code\n- `make check_code_quality` to check code quality (PEP8 basically)\n- `black .`\n- `ruff . --fix`\n\n\n## Contributing\n\n1. Fork the repository\n2. Create a feature branch (`git checkout -b feature/amazing-feature`)\n3. Commit your changes (`git commit -m 'Add amazing feature'`)\n4. Push to the branch (`git push origin feature/amazing-feature`)\n5. Open a Pull Request\n\n## Documentation\n\nFor detailed documentation, see [DOCS.md](DOCS.md).\n\n## Citation\n\nIf you use AI-CoScientist in your research, please cite:\n\n```bibtex\n@software{ai_coscientist,\n  title={AI-CoScientist: A Multi-Agent Framework for Collaborative Scientific Research},\n  author={The Swarm Corporation},\n  year={2024},\n  url={https://github.com/The-Swarm-Corporation/AI-CoScientist}\n}\n```\n\n## License\n\nMIT License - see [LICENSE](LICENSE) file for details.\n\n## Support\n\n- 📧 Email: kye@swarms.world\n- 💬 Discord: [Join our community](https://discord.gg/swarms-999382051935506503)\n- 🐦 Twitter: [@kyegomezb](https://x.com/kyegomezb)\n\n\nSource: https://github.com/The-Swarm-Corporation/AI-CoScientist",
  "tags": [
    "agents",
    "ai",
    "ml",
    "multi-agent",
    "swarms",
    "googleai",
    "health-ai",
    "ml-agents",
    "multi-agent-systems",
    "opensource",
    "python",
    "google-ai-research",
    "health-agents",
    "swarms-framework"
  ],
  "requirements": [
    {
      "package": "requests",
      "installation": "pip3 install swarms"
    }
  ],
  "usecases": [
    {
      "title": "",
      "description": ""
    }
  ],
  "userId": "12fa3db5-0635-41b6-aa8c-71ece35a18db",
  "createdAt": "2025-07-11T16:58:18.604065+00:00",
  "links": [],
  "code": "from ai_coscientist import AIScientistFramework\nimport json\nfrom loguru import logger\n\nai_coscientist = AIScientistFramework(\n    model_name=\"gemini/gemini-2.0-flash\",  # Or \"gemini/gemini-2.0-flash\" if you have access\n    max_iterations=2,  # Reduced iterations for example run\n    verbose=False,  # Set to True for detailed logs\n    hypotheses_per_generation=10,\n    tournament_size=8,\n    evolution_top_k=3,\n)\n\n# Define a research goal\nresearch_goal = \"Develop novel hypotheses for Incentivizing Reasoning Capability in LLMs via Reinforcement Learning\"\n\n# Run the research workflow\nresults = ai_coscientist.run_research_workflow(research_goal)\n\n# Output the results\nprint(\"\\n--- Research Workflow Results ---\")\nif \"error\" in results:\n    print(f\"Error during workflow: {results['error']}\")\nelse:\n    print(\"\\n--- Top Ranked Hypotheses ---\")\n    for hy in results[\"top_ranked_hypotheses\"]:\n        print(f\"- Hypothesis: {hy['text']}\")\n        print(f\"  Elo Rating: {hy['elo_rating']}\")\n        print(f\"  Score: {hy['score']:.2f}\")\n        print(\n            f\"  Reviews: {hy['reviews'][-1].get('review_summary') if hy['reviews'] else 'No reviews'}\"\n        )  # Print review summary\n        print(\n            f\"  Similarity Cluster ID: {hy['similarity_cluster_id']}\"\n        )\n        print(\n            f\"  Win Rate: {hy['win_rate']}% (Matches: {hy['total_matches']})\"\n        )\n        print(\"-\" * 30)\n\n    print(\"\\n--- Meta-Review Insights Summary ---\")\n    meta_review_summary = results[\"meta_review_insights\"].get(\n        \"meta_review_summary\", \"No meta-review summary available.\"\n    )\n    print(\n        meta_review_summary[:500] + \"...\"\n        if len(meta_review_summary) > 500\n        else meta_review_summary\n    )  # Print truncated or full summary\n\n    print(\"\\n--- Execution Metrics ---\")\n    print(json.dumps(results[\"execution_metrics\"], indent=2))\n    print(\n        f\"\\nTotal Workflow Time: {results['total_workflow_time']:.2f} seconds\"\n    )\n\n    if (\n        ai_coscientist.verbose\n    ):  # Only print full history if verbose is on, can be very long\n        print(\"\\n--- Conversation History (Verbose Mode) ---\")\n        print(\n            results[\"conversation_history\"][:1000] + \"...\"\n        )  # Print first 1000 chars of history\n\n# Save agent states (optional)\ntry:\n    ai_coscientist.save_state()\nexcept Exception as e:\n    logger.error(f\"Exception during main execution: {e}\")\n"
}