GraphRAG 🔍

An intelligent chatbot system combining Knowledge Graphs with RAG for enhanced contextual understanding.

Features • Architecture • Installation • Usage • Contributing

Table of Contents

• Introduction
• Features
• Architecture
  • Frontend
  • Backend
  • Knowledge Graph Structure 🔍
• Installation
  • Prerequisites
  • Setup
• Usage
• Contributing
• Acknowledgements
• License

Introduction

GraphRAG enhances traditional chatbots by leveraging Knowledge Graphs, combining structured and unstructured data into a comprehensive Graph Database. This approach enriches the standard RAG (Retrieval-Augmented Generation) system with relationship-aware context, enabling more accurate and contextual responses to complex queries.

Features

• 🔍 Advanced Query Understanding: Utilizes graph-based context for better query comprehension
• 🧠 Intelligent Response Generation: Combines GPT-4 with graph-based knowledge retrieval
• 📊 Knowledge Graph Integration: Neo4j-powered data relationships
• 🎯 Contextual Awareness: Maintains context through graph relationships
• 💨 Real-time Processing: Fast response generation with optimized retrieval
• 🎨 Modern UI: Sleek interface built with Next.js and TailwindCSS

Architecture

For a detailed understanding of the system architecture, data flow, and component interactions, please see our detailed architecture documentation.

        graph TB
        subgraph Frontend["Frontend (Next.js)"]
            UI[Chat Interface]
            State[React State Management]
            API[Axios API Client]
        end

        subgraph Backend["Backend (FastAPI)"]
            FastAPI[FastAPI Server]
            SearchEngine[Neo4j Search Engine]
            LLM[GPT-4 Integration]
            QueryEmbedding[Query Embedding Service]
        end

        subgraph DataSources["Data Sources"]
            SQL[(SQLite Database)]
            PDF[PDF Documents]
        end

        subgraph EmbeddingProcess["Embedding Process"]
            ArtistEmbed[Artist Embedding]
            AgentEmbed[Sales Agent Embedding]
            DocEmbed[Document Embedding]
            TextSplitter[Text Splitter]
        end

        subgraph Database["Knowledge Graph (Neo4j)"]
            Neo4j[(Neo4j Database)]
            Artists[Artist Nodes]
            Agents[Sales Agent Nodes]
            Documents[Document Nodes]
            Relationships{Graph Relationships}
        end

        %% Frontend Flow
        UI --> State
        State --> API
        API -->|HTTP POST| FastAPI

        %% Data Source Processing
        SQL -->|Extract Data| ArtistEmbed
        SQL -->|Extract Data| AgentEmbed
        PDF -->|Load| TextSplitter
        TextSplitter -->|Chunks| DocEmbed

        %% Embedding Process
        ArtistEmbed -->|Embedded Data| Artists
        AgentEmbed -->|Embedded Data| Agents
        DocEmbed -->|Embedded Data| Documents

        %% Query Processing
        FastAPI -->|User Query| QueryEmbedding
        QueryEmbedding -->|Search Similar| SearchEngine
        SearchEngine -->|Graph Query| Neo4j
        SearchEngine -->|Generate Response| LLM

        %% Database Relationships
        Artists -->|MANAGED| Relationships
        Agents -->|MENTIONS| Relationships
        Documents -->|RELATED_TO| Relationships

        %% Response Flow
        LLM -->|Structured Response| FastAPI
        FastAPI -->|JSON| API
        API -->|Update| State
        State -->|Render| UI

        classDef frontend fill:#6366f1,color:#fff,stroke:#4338ca
        classDef backend fill:#16a34a,color:#fff,stroke:#15803d
        classDef database fill:#0891b2,color:#fff,stroke:#0e7490
        classDef process fill:#8b5cf6,color:#fff,stroke:#7c3aed
        classDef datasource fill:#ea580c,color:#fff,stroke:#c2410c

        class UI,State,API frontend
        class FastAPI,SearchEngine,LLM,QueryEmbedding backend
        class Neo4j,Artists,Agents,Documents,Relationships database
        class ArtistEmbed,AgentEmbed,DocEmbed,TextSplitter process
        class SQL,PDF datasource

Loading

Frontend

• Framework: Next.js with TypeScript
• Styling: TailwindCSS with custom animations
• Components:
  • Chat interface with real-time updates
  • Markdown rendering for formatted responses
  • Loading states and animations

Backend

• API: FastAPI server handling query processing
• LLM Integration: OpenAI GPT-4 for response generation
• Embedding: OpenAI Ada-002 for text embeddings
• Database: Neo4j for graph storage and querying

Knowledge Graph Structure 🔍

Visual representation of the Neo4j Knowledge Graph structure showing entity relationships

Node Types

• 🎨 Artists: Music artists with revenue data and embeddings
• 👥 Sales Agents: Representatives managing artists
• 📄 Documents: Embedded text chunks from documentation

Relationships

• MANAGED: Links sales agents to their managed artists
• MENTIONS: Connects documents to referenced entities
• RELATED_TO: Creates semantic connections between similar documents

The graph structure enables:

• Contextual information retrieval
• Relationship-aware querying
• Semantic similarity search
• Multi-hop reasoning capabilities

Installation

Prerequisites

• Node.js (v18+)
• Python (v3.8+)
• Neo4j Database
• OpenAI API Key

Setup

1. Clone the repository

    bash
    git clone https://github.com/yourusername/graphrag.git
    cd graphrag

2. Frontend Setup

    cd src/frontend
    npm install

3. Backend Setup

    cd src/backend
    python -m venv venv
    source venv/bin/activate # or venv\Scripts\activate on Windows
    pip install -r requirements.txt

4. Neo4j Setup

• Install Neo4j Desktop
• Create a new database
• Set password to '12345678' or update in configuration
• Start the database

5. Environment Variables Create .env file in backend directory:

    OPENAI_API_KEY=your_api_key_here
    NEO4J_URI=bolt://localhost:7687
    NEO4J_USERNAME=neo4j
    NEO4J_PASSWORD=12345678

Usage

1. Start the Backend

    cd src/backend
    uvicorn generateResponse:app --reload

2. Start the Frontend

    cd src/frontend
    npm run dev

3. Access the Application Open http://localhost:3000 in your browser

Contributing

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

Acknowledgements

• OpenAI for GPT-4 and embeddings
• Neo4j for graph database
• Vercel for Next.js
• All contributors and supporters

License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Made with ❤️ by @hrithikkoduri