TL;DR
Yuxi is a multi-tenant agent harness platform integrating knowledge bases and knowledge graphs, combining RAG search and multi-agent orchestration to support enterprise intelligent search and task automation.
Key features
Multi-tenant support: Separated knowledge bases, models, and permission management per organization
Knowledge base + knowledge graph: Combines Milvus-based vector search with Neo4j-based graph reasoning
Multi-agent orchestration: LangGraph-based execution of complex workflows combining skills, MCP, sub-agents, and sandbox tools
Document parsing: Advanced document understanding using MinerU, PaddleX, and RapidOCR
RAG search: LightRAG-style retrieval-augmented generation pipeline
Docker Compose deployment: Easy installation and operation
When to use it
When building an enterprise internal knowledge base and needing an LLM-based Q&A system
In multi-tenant environments where multiple departments or teams need independent knowledge bases and agents
For task automation scenarios beyond simple Q&A that require complex reasoning, tool use, and multi-step tasks
When relationship-based search and reasoning using knowledge graphs are needed