Kusanagi-AI: Free & Open-Source Local AI Toolkit

About This Project

Kusanagi-AI was developed to address the growing need for accessible and privacy-conscious AI solutions within the research community. Our mission is to provide a free, open-source toolkit that enables researchers, especially those in Physics and Material Science, to leverage advanced AI capabilities directly on their personal computers. By focusing on local execution, Kusanagi-AI ensures complete data privacy and eliminates reliance on cloud services, making sophisticated AI analysis available without specialized hardware or extensive technical expertise. This project is a testament to the power of local AI, offering a controlled environment for deep learning and practical application.

Technology Stack

Python

The entire frontend and application logic are developed in Python, ensuring readability and a vast ecosystem of libraries.

Ollama

Powers local LLM inference, allowing Kusanagi-AI to run models like Llama3 and Gemma efficiently without cloud dependencies.

MXBAI Embeddings

Utilizes the `mxbai-embed-large` model for high-quality document embeddings, crucial for RAG and semantic search.

Features

• Local & Private: All operations are 100% offline, guaranteeing your research data remains secure.
• Efficient Local LLM Inference: Run up to three LLMs concurrently on a decent laptop.
• Research Assistant (Orochimaru): A flagship application for academic use with a Python-based frontend.
• Engage in RAG with your PDF documents for in-depth analysis.
• Generate summaries and peer reviews of research papers.
• Experimental Chatbots: A collection of scripts for exploring different AI models.
• AI Visualizer: Tools for visualizing AI-related data and model outputs.

Getting Started

git clone https://github.com/prathameshnium/Kusanagi-AI.git
cd Kusanagi-AI

pip install -r requirements.txt

ollama pull mxbai-embed-large
ollama pull llama3

python Orochimaru_Local_Research_Assistent.py