xerrors/Yuxi

结合知识库、知识图谱管理的 多租户 Agent Harness 平台。 An agent harness that integrates a LightRAG knowledge base and knowledge graphs. Build with LangChain + Vue + FastAPI, support DeepAgents、MinerU PDF、Neo4j 、MCP.

What it solves

Yuxi provides a multi-tenant platform for enterprises to transform their internal knowledge into a format that AI agents can retrieve, reason over, and deliver. It bridges the gap between raw corporate data and actionable AI intelligence by integrating knowledge bases and knowledge graphs into a unified workspace.

How it works

The platform combines several advanced AI techniques into a single workflow:

  • RAG & Knowledge Graphs: It uses RAG (Retrieval-Augmented Generation) and Milvus-based knowledge graphs to allow agents to perform complex reasoning over stored data.
  • Agent Orchestration: Built on LangGraph, it supports multi-agent orchestration, allowing users to create agents that can use "Skills," MCP (Model Context Protocol), sub-agents, and sandbox tools.
  • Multi-tenant Management: Administrators can configure knowledge bases, models, and permissions through a centralized workbench.
  • User Interface: Users interact with agents via a ChatGPT-like interface to receive answers with cited sources and knowledge graph-based reasoning.

Who it’s for

It is designed for enterprises and organizations that need a secure, multi-tenant environment to deploy intelligent knowledge bases and develop AI agents capable of handling complex corporate data.

Highlights

  • Integrated Tech Stack: Combines FastAPI, LangGraph, Milvus, and Neo4j for robust data handling and agent logic.
  • Advanced Document Parsing: Utilizes MinerU, PaddleX, and RapidOCR for high-quality document processing.
  • Multi-Agent Capabilities: Supports the deployment of agents with specialized skills and sandbox tool integration.
  • Knowledge Graph Reasoning: Goes beyond simple vector search by incorporating graph-based reasoning for more accurate answers.