Welcome to graph-nd documentation!

A Python package for building powerful end-to-end agentic GraphRAG systems with a simple, intuitive API.

• Zero-to-GraphRAG in under 5 minutes - no prior graph expertise required.

• Built with intent to extend to production - not just a demo tool.

• Prioritizes support for mixed (structured + unstructured) data sources

• Out-of-the-Box GraphRAG Agents - Automatically builds extensible GraphRAG agents customized to your graph schema

Contents:

• Getting Started
• Examples
• API Reference
  • graph_nd package
• Contributing

Features:

• Knowledge Graph Creation & Data Integration: Easily map mixed data sources to knowledge graphs (KGs):

• tables from CSVs, RDBMS, and Data Warehouses

• text from PDFs and other files via KG driven entity extraction

• other data from your own custom workflows with optimized node and relationship loading

• Neo4j Persistence:

• Store KGs in Neo4j graph databases for reliability & scale

• Leverage graph database traversals & indexing for low-latency retrieval

• Out-of-the-Box GraphRAG Agents

• Automatically builds Langgraph ReAct agents with tools and system prompts customized to your graph schema and use case

• Enables agent customization through the addition of other user defined tools and prompt context, either locally or via MCP

• Coming Soon - Agent factories for creating agents in multiple other frameworks and SDKs