Yuxi-Know is a knowledge graph question-answering system built upon large language models and Retrieval-Augmented Generation (RAG). It targets developers and researchers seeking a flexible, multi-model platform for knowledge retrieval, integrating document-based RAG with Neo4j knowledge graphs. The system offers broad LLM compatibility and extensibility for custom agents.

How It Works

The system leverages Langgraph for agentic workflows, FastAPI for the backend API, VueJS for the frontend, and Neo4j for knowledge graph storage. It supports various LLMs (OpenAI, Ollama, vLLM, domestic platforms) and embedding models, including DeepSeek-R1. Document ingestion involves converting PDFs, TXTs, and MDs into text, then embedding them for vector storage. Knowledge graphs are ingested in a jsonl format with {"h": "head", "t": "tail", "r": "relation"} triples.

Quick Start & Requirements

• Install/Run: docker compose -f docker/docker-compose.dev.yml --env-file src/.env up --build
• Prerequisites: API keys for chosen LLM and embedding services (e.g., SILICONFLOW_API_KEY, OPENAI_API_KEY). Requires Docker.
• Setup: Minimal setup involves creating .env from .env.template and providing API keys.
• Links: Project GitHub

Highlighted Details

• Supports multiple LLM providers via API, including local deployments through OpenAI-compatible endpoints (vLLM, Ollama).
• Integrates web search capabilities (requires TAVILY_API_KEY) and tool-use for agents (WIP).
• Offers a configurable knowledge graph visualization and management interface.
• Allows adding custom LLM and embedding models via models.yaml.

Maintenance & Community

The project is actively developed with recent updates in March 2025. It acknowledges contributors and provides links to GitHub issues and stars.

Licensing & Compatibility

The project is licensed under the MIT license. This license permits commercial use and integration with closed-source projects.

Limitations & Caveats

The agent functionality is marked as "Work In Progress" (WIP) and not a formal release. The OneKE tool for automatic knowledge graph creation has been removed due to poor performance, recommending external graph creation. The system currently does not support querying multiple knowledge graphs simultaneously.