Through this open-source repo and article, we are pleased to demonstrate the power of multi-knowledge graph multi-agent systems for traversing documents with an example of a legal regulatory document. Particularly useful for legal contracts and manufacturing manuals, this system allows you to recursively retrieve different clauses, references to other sections, footnotes and definitions throughout the legal document. This closely mimics the unique way that legal documents (and many types of manuals) set out information in specific structured ways that have references throughout. For example, if a section refers to ‘Paragraph 9.1.’ , it can refer to a document graph and bring in Paragraph 9.1. Or if the clause has a footer note that says ‘refer to paragraph 7.3 and 7.4’, it can retrieve the footer note, and then retrieve the paragraphs. Or if the clauses has a term that is defined in the definitions sheet, it can bring in the definitions from the definitions graph. This is incredibly useful in any RAG system that has documentation that refers to other parts of the document (intra-document hierarchies) or other documents in the context of manufacturing manuals, etc (inter-document hierarchies). https://medium.com/enterprise-rag/legal-document-rag-multi-graph-multi-agent-recursive-retrieval-through-legal-clauses-c90e073e0052