graphrag

GraphRAG — Graph-based Retrieval Augmented Generation

Scripts demonstrating knowledge graph construction and graph-based QA using LangChain's NetworkxEntityGraph and GraphQAChain, with Groq as the LLM backend (via the OpenAI-compatible endpoint).

Files

File	Description
graphrag_langchain_spicejet.py	Builds a flight-network graph from CSV and answers route queries
graphrag_langchain_test.py	Minimal example: builds a graph from plain text sentences and runs QA

Setup

pip install langchain langchain-openai langchain-community networkx pandas
export GROQ_API_KEY=your_key_here

Running

# Requires data/flight_network_spicejet.csv
python graphrag_langchain_spicejet.py

# Self-contained test with hardcoded text
python graphrag_langchain_test.py

Architecture

Text / CSV  →  GraphIndexCreator / manual edges  →  NetworkxEntityGraph
                                                          ↓
                                                    GraphQAChain (LLM)
                                                          ↓
                                                    Natural language answer

Notes

• Both scripts use ChatOpenAI pointed at https://api.groq.com/openai/v1 — a standard pattern for using Groq via the OpenAI SDK.
• Reference: LangChain GraphRAG article