🔬 Advanced RAG Comparison and Analysis

Welcome to the Advanced RAG Comparison and Analysis Lab!

This Streamlit-based sandbox allows AI enthusiasts, students, and researchers to explore the internals of Retrieval Augmented Generation (RAG).

Stay tuned! More advanced RAG techniques & features are underway and will be added soon.

🚀 Highlights

https://rag-by-mk.streamlit.app/

• Interactive Playground to tweak and test RAG components.
• Side-by-Side Comparison of various indexing, retrieval, and LLM configurations.
• Detailed Metadata Inspection to understand each processing stage.

✨ Core Features

• Indexing Backbones: Vector Store & Knowledge Graph indexes
• LLM Selection: Choose models like Llama3, DeepSeek-r1-distill-llama-70b (via Groq API)
• Retrieval Techniques: Vector-only, Sparse (BM25), and Hybrid search
• Query Optimization: Enable Query Transformations like HyDE
• Re-ranking: Use sentence transformer models to refine results
• LLM-as-a-Judge: Evaluate output coherence and relevance

🛠️ Tech Stack

• Frontend: Streamlit
• Framework: LlamaIndex
• LLMs: Groq API (Llama3, DeepSeek Coder)
• Embeddings: FastEmbed
• Rerankers: Sentence Transformers
• Sparse Retrieval: BM25
• Document Parsing: PyMuPDF, python-docx, openpyxl, Markdown

📁 Folder Structure

LLM/
├── data_for_rag/         # Upload documents here
├── persisted_index/      # Stores indexes and document states
├── ui/                   # Streamlit UI components
├── utils/                # Helper functions
├── workflow/             # RAG logic (rag_workflow.py)
├── config.py             # App configuration
├── main.py               # Entry-point script
└── requirements.txt      # Dependencies

🚀 Setup Instructions

Prerequisites

• Python 3.9+
• Groq API Key

Installation

git clone https://github.com/MohanKrishnaGR/LLM.git
cd LLM
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r LLM/requirements.txt

Run

streamlit run LLM/main.py

How to Use

1. Add Documents: Place files in LLM/data_for_rag/.
2. Configure Sidebar: Choose LLM, index type, chunk size, etc.
3. Build Index: Click 'Create / Update Index'.
4. Experiment: Adjust retrieval & transformation settings.
5. Query & Analyze: Run queries and view detailed metadata.

👤 About the Developer

Built by Mohan Krishna G R, passionate about GenAI and RAG systems.

• LinkedIn
• Portfolio

---