What is Yuxi-Know
Yuxi-Know is an intelligent question-answering platform built on large model RAG knowledge bases and knowledge graph technology. Yuxi-Know supports a variety of knowledge base file formats (such as PDF, TXT, MD, Docx), and supports converting file content into vector storage for fast retrieval. Yuxi-Know integrates Neo4j-based knowledge graph question-answering capabilities and can handle complex knowledge relationship queries. The platform supports multi-model adaptation, including OpenAI, mainstream domestic large models, and locally deployed vllm, ollama, etc. Yuxi-Know has agent extension capabilities, supporting developers in writing custom agent code to further enhance the system's flexibility and functionality.
Main Features of Yuxi-Know
- Multi-Model Support: Supports multiple large model platforms, including OpenAI, mainstream domestic large model platforms, and locally deployed vllm and ollama.
- Flexible Knowledge Base Management: Supports uploading documents in various formats such as PDF, TXT, MD, Docx, etc.
- Knowledge Graph Integration: Neo4j-based knowledge graph question-answering capability, supporting users in uploading knowledge graph files in jsonl format.
- Agent Extension: Supports users in writing custom agent code to further extend the system's functionality.
- Reasoning Model Support: SupportsDeepseek-R1and other reasoning models, providing a more intelligent question-answering experience.
- Web Retrieval: Supports web content retrieval and display.
- Visual Configuration: Provides rich visual configuration features, making it convenient for users to manage and optimize system settings.
Technical Principles of Yuxi-Know
- RAG Architecture: Based on a vector model, the text in the knowledge base is converted into vectors and stored in a vector database. When a user asks a question, the system retrieves the document fragments most relevant to the question from the vector database. The retrieved document fragments and the user's question are input together into a large language model to generate an accurate answer.
- Knowledge Graph Technology: Neo4j serves as the storage and query engine for the knowledge graph. The knowledge graph stores knowledge in the form of nodes and relationships, and users perform complex knowledge relationship queries based on the graph. The construction and management of the knowledge graph are completed through a web interface, allowing users to conveniently upload and update graph data.
- Multi-Model Adaptation: Based on configuration files (such as models.yaml), the system supports API calls to multiple large model platforms. Users select different models as needed, and the system automatically adapts and calls the corresponding API.
- Vector Model and Reranking Model: The vector model converts text into vectors and supports multiple vector models, such as BAAI/bge-m3, etc. The reranking model optimizes the order of retrieval results and improves the accuracy and efficiency of question answering.
- Frontend and Backend Technology Stack: The frontend is based on VueJS, providing a user-friendly interactive interface. The backend is based on FastAPI, providing efficient server-side processing capabilities. The database uses Milvus-Standalone as the vector database and Neo4j as the knowledge graph database.
- Docker Containerized Deployment: Docker is used for containerized deployment, facilitating the rapid setup and management of development and production environments. Supports automatic updates and background operation in the development environment.
Yuxi-Know Project Address
- GitHub Repository：https://github.com/xerrors/Yuxi-Know
Application Scenarios of Yuxi-Know
- Enterprise Knowledge Management: Build enterprise knowledge bases, enabling employees to quickly look up information and improve work efficiency.
- Education: Assist online learning, providing students with intelligent tutoring and knowledge answers.
- Customer Service: As an intelligent customer service, quickly answer customer inquiries and improve customer satisfaction.
- Healthcare: Provide medical knowledge queries and preliminary patient consultations, assisting medical decision-making.
- Scientific Research and Academia: Support literature retrieval and knowledge graph construction, facilitating scientific research work.