SOTA Rivalry Tracker: Lightweight Graph RAG

A specialized Graph Retrieval-Augmented Generation (RAG) system built to track and analyze the competitive landscape of State-of-the-Art (SOTA) AI models in 2026. This project utilizes NetworkX for knowledge graph management and Groq (Llama 3.3) for intelligent reasoning over retrieved relationships.

Key Features

• Lightweight Graph Construction: Efficiently builds a knowledge graph from structured data sources,
• Dynamic Retrieval: Retrieves relevant relationships and insights based on user queries,
• Intelligent Reasoning: Uses Groq to analyze retrieved data and generate insightful responses
• SOTA Model Tracking: Focuses on tracking the latest advancements in AI models, their performance metrics, and competitive dynamics.
• User-Friendly Interface: Provides an intuitive interface for users to interact with the system and explore the competitive landscape of SOTA models.

Installation

1. Clone the repository:

git clone https://github.com/husseinwork6/sota-graph-rag.git
cd sota-graph-rag

2. Set up a virtual environment:

python -m venv venv
.\venv\Scripts\activate  # Windows

3. Install dependencies:

pip install streamlit groq networkx python-dotenv

4. configure environment variables:

# Create a .env file in the root directory and add your Groq API key
GROQ_API_KEY=your_groq_api_key_here

Usage

Run the Streamlit app:

streamlit run app.py

Knowledge Base

The system is currently grounded in a data.json file containing over 40+ verified 2026 benchmarks, including:

• Reasoning Models: OpenAI o1/o3, DeepSeek-R1.

• Open Weights: Llama 4 Preview, Mistral Large 2.

• Benchmarks: MMLU, HumanEval, GPQA, and SWE-bench.