I created the RAG‑Benchmark‑System repository to provide a standardized benchmark framework for evaluating Retrieval‑Augmented Generation architectures using enterprise‑style documents.
The repository is named RAG Benchmark System.
It is implemented in Python.
The repository has 1 star.
It was created in 2026-08.
The last update was in 2026-10.
I chose a modular architecture for reuse and extension.
Vector retrieval with ChromaDB is used as the vector store.
Graph RAG support uses Neo4j‑style graph construction.
Multimodal/PageIndex‑aware retrieval workflow is included.
In this repository, I’ve implemented and benchmarked seven different approaches: Simple RAG, Fusion RAG, Self RAG, Adaptive RAG, Agentic RAG, Graph RAG, and Multimodal RAG / PageIndex‑style retrieval.
from rags.simple_rag.simple_rag import SimpleRAG
from rags.fusion_rag.fusion_rag import FusionRAG
simple = SimpleRAG()
fusion = FusionRAG()
question = "What is Hemant Sharma's employee ID?"
print(simple_result["answer"])\nprint(fusion_result["answer"])