ProPEX-RAG Enhanced GraphRAG using Prompt Driven Prompt Execution
-
Updated
Oct 18, 2026 - Python
ProPEX-RAG Enhanced GraphRAG using Prompt Driven Prompt Execution
Memgraph storage backends (KV, Vector, DocStatus) for LightRAG
Entity extraction, community detection, local and global search, Personalized PageRank, Text2Cypher, temporal graphs and evaluation — GraphRAG built from scratch in Python, runnable offline.
A hands-on RAG learning repository exploring Naive RAG, Advanced RAG, GraphRAG, and future retrieval architectures.
Hybrid GraphRAG helps Agent Zero understand how your notes and documents connect, so it can trace “why,” surface related context, and keep long-running work coherent across large knowledge sets.
To associate your repository with the graphrag-rag topic, visit your repo's landing page and select "manage topics."