NexusBase: Enterprise RAG Architecture

NexusBase is a production-grade Retrieval-Augmented Generation (RAG) system designed for enterprise-scale knowledge management. It leverages a modern agentic workflow using LangGraph to handle complex document reasoning, multi-step retrieval, and structured responses.

🏗️ Architecture

The system follows a modular, containerized architecture designed for high availability and low latency.

graph TD
    User([User/Frontend]) <--> Dashboard[Next.js Dashboard]
    Dashboard <--> API[FastAPI Backend]
    
    subgraph "Agentic Pipeline (LangGraph)"
        API --> Router{Semantic Router}
        Router -- Semantic Query --> Retreiver[pgvector Retrieval]
        Router -- Direct Knowledge --> Summarizer[Summarization Node]
        Retreiver --> Evaluator{Response Evaluator}
        Evaluator -- Loop Back --> Retreiver
        Evaluator -- Validated --> Output[Structured Output]
    end
    
    Output --> API
    API --> DB[(PostgreSQL + pgvector)]

Loading

🚀 Quick Start (Local Deployment)

NexusBase is fully containerized with Docker. To spin up the entire stack (Dashboard, API, Database, and Workers), run the following command:

Prerequisites

• Docker & Docker Compose
• OpenAI API Key (or local LLM configuration in .env)

Deployment

# Clone the repository (if not already local)
# git clone <your-repo-url>
# cd NexusBase

# Initialize environment variables
cp .env.example .env

# Launch the stack
docker compose up --build

The services will be available at:

• Dashboard: http://localhost:3000
• Backend API: http://localhost:8000
• Database: localhost:5432

🛠️ Tech Stack

• Frontend: Next.js 14, TypeScript, Tailwind CSS, Lucide React.
• Backend: Python 3.11, FastAPI, Pydantic.
• Orchestration: LangGraph, LangChain.
• Vector Store: PostgreSQL with pgvector extension.
• DevOps: Docker, Docker Compose, GitHub Actions.

📄 License

This project is licensed under the MIT License.