https://github.com/nitin-sagar-b/ratio-core
A modular multi-model AI framework demonstrating advanced techniques in semantic knowledge transfer, context management, and collaborative intelligence across diverse language models.
https://github.com/nitin-sagar-b/ratio-core
Last synced: 2 months ago
JSON representation
A modular multi-model AI framework demonstrating advanced techniques in semantic knowledge transfer, context management, and collaborative intelligence across diverse language models.
- Host: GitHub
- URL: https://github.com/nitin-sagar-b/ratio-core
- Owner: Nitin-Sagar-B
- License: mit
- Created: 2025-03-26T04:10:17.000Z (7 months ago)
- Default Branch: main
- Last Pushed: 2025-03-26T05:14:31.000Z (7 months ago)
- Last Synced: 2025-03-26T05:22:31.805Z (7 months ago)
- Language: Python
- Size: 121 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# CoRE: Collaboration of Role-based Experts 🤖
## 🌟 Project Overview
CoRE (Collaboration of Role-based Experts) is an innovative AI system that demonstrates a novel approach to collaborative reasoning by leveraging multiple specialized language models. Built on the RaTiO (Reasoning through Interaction of Experts) framework, CoRE showcases how smaller, focused AI models can work together to generate high-quality, nuanced responses.
## Dashboard Demo: https://ratio-core.streamlit.app/
The dashboard UI can be accessed using the above link but the local model functionality won't be performed online, hence follow the instructions below to complement the application on your device.
## 🚀 Key Features
- **Multi-Expert Collaboration**: Combine insights from different specialized AI models
- **Dynamic Expert Selection**: Intelligently route queries to most relevant experts
- **Semantic Knowledge Transfer**: Build and reuse insights across interactions
- **Tiered Context Management**: Maintain conversation history with intelligent compression
- **Interactive Visualization**: Gain insights into expert contributions
## 🔧 System Architecture
CoRE integrates several cutting-edge components:
- Expert Models: Gemma, LLaMA, Phi, Qwen, Gemini
- Context Management System
- Knowledge Transfer Hub
- Expert Gating Mechanism
- Streamlit-based User Interface
## 💻 Technology Stack
- Python 3.8+
- Streamlit
- Langchain
- Ollama
- SentenceTransformer
- Plotly
- Google Generative AI
- Scikit-learn
## 🛠 Installation
### Prerequisites
- Python 3.8+
- Ollama with pre-downloaded models
- Gemini API Key (optional)
### Setup
```bash
# Clone the repository
git clone https://github.com/yourusername/CoRE.git
cd CoRE
# Install dependencies
pip install -r requirements.txt
# Run the application
streamlit run v2.py
```
## 🎮 Usage
1. Launch the Streamlit application
2. Enter your query
3. Choose expert selection mode (Automatic/Manual)
4. Configure expert roles (optional)
5. Select an aggregator model
6. Process the query and explore results!
## 📊 Visualization Features
- **Expert Relevance Radar Chart**: Understand query complexity
- **Response Flow Diagram**: Visualize expert contributions
## 🔬 Research Insights
CoRE demonstrates:
- Efficient use of smaller language models
- Collaborative reasoning approach
- Semantic knowledge transfer
- Low-resource AI processing
## 🚧 Future Roadmap
- Parallel processing support
- Enhanced knowledge management
- User feedback integration
- Domain-specific configurations
- Multi-modal input support
## 👥 Collaborators
This project is developed and maintained by:
- [Boyeena Nitin Sagar](https://github.com/Nitin-Sagar-B)
- [Begari Susheel Kumar](https://github.com/specialsusheel)
- [Kuchuru Sainath Reddy](https://github.com/sainath-03)
## 📜 License
This project is licensed under the **MIT License**. See the [LICENSE](LICENSE) file for details.
## 🤝 Contributing
Contributions are welcome! Please read our contributing guidelines before getting started.
[](https://github.com/Nitin-Sagar-B/RaTiO-CoRE/graphs/contributors)
## Acknowledgements
- Inspired by RaTiO Framework
- Powered by open-source AI technologies
---
**Star ⭐ the repo if you find it interesting!**