🚀 Think-on-Graph 3.0: Efficient and Adaptive LLM Reasoning on Heterogeneous Graphs via Multi-Agent Dual-Evolving Context Retrieval

A factory for building advanced RAG (Retrieval-Augmented Generation) pipelines, including:

• Standard RAG implementations
• GraphRAG architectures
• Multi-modal RAG systems

🌟Features

• Modular design for easy customization
• Support for various knowledge graph backends
• Integration with multiple LLM providers
• Configurable pipeline components

Installation

pip install -e .

Usage

bash run.sh naive_rag/graph_rag/mm_rag

or

python main.py --config examples/graphrag/config.yaml

Examples

See the examples/ directory for sample configurations and usage.

Roadmap

✅ Implemented Features

• Vector RAG (基于Qdrant实现)
• Graph RAG (基于Neo4j实现)
• Multi-modal RAG (基于Neo4j实现文本和图像向量存储与检索)
• Lightweight SQLite Cache (轻量级缓存方案)

🚧 Planned Features

• ReAct QueryEngine (交互式查询引擎)
• Query Engineering:
  • Query Rewriting (查询重写)
  • Sub-Questions (子问题分解)
• Agentic RAG (智能工具选择优化性能)

🙏 Acknowledgements

This project draws inspiration from and gratefully acknowledges the contributions of the following open-source project:

• llama-index

• llama-factory

• Qdrant

• Neo4j

  📖 Citation

@misc{wu2025ToG-3,title={Think-on-Graph3.0: EfficientandAdaptiveLLMReasoningonHeterogeneousGraphsviaMulti-AgentDual-EvolvingContextRetrieval}, author={XiaojunWu, CehaoYang, XueyuanLin, ChengjinXu, XuhuiJiang, YuanliangSun, HuiXiong, JiaLi, JianGuo},year={2025},eprint={2509.21710},archivePrefix={arXiv},primaryClass={cs.CL},url={https://arxiv.org/abs/2509.21710}, }

⭐ Star History

⭐ 如果这个项目对您有帮助，动动小手点亮Star吧！