Report Abuse

Basic Information

graph-rag-agent is a developer-focused project that integrates GraphRAG, LightRAG and neo4j-llm-graph-builder to enable knowledge graph construction, graph-backed search, and retrieval-augmented generation (RAG) workflows. The repository targets building and searching knowledge graphs with Neo4j and combining graph retrieval results with language model inference. It integrates DeepSearch techniques to enable private-domain RAG inference and includes a custom evaluation framework designed specifically for assessing GraphRAG-style systems. The project is intended for researchers and engineers who want to experiment with, reproduce, or extend graph-centric RAG approaches and to evaluate retrieval and generation performance within controlled or private datasets.

Links

Categorization

App Details

Features
Integrates GraphRAG, LightRAG and neo4j-llm-graph-builder to support automated knowledge graph construction and graph-based retrieval. Implements graph search and retrieval pipelines that can be combined with LLM inference. Incorporates DeepSearch methods to enable private or on-premise RAG inference over domain data. Provides a bespoke evaluation framework tailored to benchmark GraphRAG variations and measure retrieval and generation outcomes. Emphasizes modular integrations for ingestion, Neo4j graph building, retrieval orchestration, and evaluation to support research and prototyping of graph-backed RAG systems.
Use Cases
This repository helps developers and researchers set up, run, and evaluate graph-centered RAG systems by consolidating integrations for Neo4j-based graph construction, graph retrieval, and LLM inference. It enables private-domain inference via DeepSearch so organizations can perform RAG without exposing sensitive data externally. The included evaluation framework supports comparative benchmarking of GraphRAG and related approaches, helping teams quantify retrieval effectiveness and downstream generation quality. By bringing together established graph-RAG components and an assessment toolset, the project reduces integration effort for prototyping, reproducing experiments, and extending graph-backed RAG models in research or controlled deployments.

Please fill the required fields*