Endee RAG AI Assistant

A production-ready Retrieval-Augmented Generation (RAG) system built using:

• Node.js (Backend)
• React (Frontend)
• Ollama (LLM + Embeddings)
• Endee (Vector Database)

---

🚀 Project Overview

This project demonstrates a fully functional AI assistant powered by semantic search and vector retrieval.

The system allows users to:

1. Store large knowledge text into a vector database
2. Automatically chunk and embed the text
3. Retrieve relevant context using semantic similarity
4. Generate accurate answers using a local LLM (Ollama)

This simulates real-world AI assistant architecture used in modern production systems.

---

🧠 Problem Statement

Large Language Models (LLMs) do not retain persistent memory.
This project solves that problem using:

• Vector embeddings
• Efficient similarity search
• Context compression
• Retrieval-Augmented Generation (RAG)

---

🏗 System Architecture

User → React UI → Node Backend →
Embedding (Ollama) → Store in Endee →
Query → Semantic Search → Context Compression → LLM → Response

---

🔎 How Endee Is Used

Endee acts as the vector database.

Workflow:

• Text is split into chunks
• Each chunk is converted into embeddings
• Embeddings are stored in Endee
• During queries, similarity search retrieves top relevant chunks
• Context is passed to LLM for final answer generation

Endee enables:

• Fast semantic retrieval
• Scalable memory storage
• Efficient vector indexing

---

🛠 Tech Stack

Backend:

• Node.js
• Express.js
• Axios
• UUID

Frontend:

• React
• CSS

AI:

• Ollama (llama3 + nomic-embed-text)
• Endee vector database

---

🎬 Demo Video

Video.Project.1.mp4

⚙️ Setup Instructions

Before you begin, ensure you have the following installed on your machine:

• Node.js (v16 or higher)
• Docker Desktop (Must be actively running)
• Ollama (For running local LLMs)

1. Clone the Repository

Clone the project to your local machine and navigate into the directory:

git clone https://github.com/Harsh2o/ENDEE-RAG-AI-ASSISTANT.git
cd ENDEE-RAG-AI-ASSISTANT

2. Start External Services (Database & LLM)

This application requires the Endee vector database and Ollama to be running before starting the servers.

• Start Docker Desktop: Ensure the Docker daemon is running.
• Start Endee Vector Database: Ensure your Endee container is running and accessible on port 8080.
• Start Ollama: Start the Ollama server (accessible on port 11434) and ensure your Llama model is pulled.

3. Install Dependencies

Install the required Node packages for the root project, as well as the frontend and backend.

npm install

(Note: If required by your folder structure, navigate to the /frontend and /backend folders to run npm install inside them as well).

4. Run the Application

Start both the backend and frontend development servers concurrently:

npm run dev

• The Backend server will start on http://localhost:5000
• The Frontend Vite application will start on http://localhost:5174 (port may vary based on availability)

5. Usage

1. Open your browser and navigate to the frontend URL (e.g., http://localhost:5174).
2. Store Knowledge: Input text or documents into the interface to save them into the Endee vector database locally.
3. Ask Questions: Ask questions through the chat interface. The backend will retrieve related context from Endee and generate a precise answer using the local Llama model.