⚛ Ragnova (AI Chat Application with RAG)

🧠 Overview

This project demonstrates the integration of Google Gemini 2.0 Flash with a Retrieval-Augmented Generation (RAG) pipeline using Streamlit and MySQL (XAMPP).
It enables real-time chat, document-based Q&A, and keyword search through a clean and interactive UI.

---

🧩 Technologies Used

Component	Description
AI Model	Google Gemini 2.0 Flash (via google-generativeai API)
Embeddings Model	Sentence Transformers – all-MiniLM-L6-v2
Programming Language	Python 3.10 +
Frameworks / Libraries	Streamlit, scikit-learn, sentence-transformers, python-dotenv
Database	MySQL (XAMPP)

---

⚙️ Setup and Installation

Step 1 – Create Virtual Environment

python -m venv venv
venv\Scripts\activate        # Windows
# or
source venv/bin/activate     # macOS/Linux

Install dependencies:

pip install -r requirements.txt

---

Step 2 – Add Gemini API Key

Create a file named .env in the root project folder and include:

GEMINI_API_KEY=your_gemini_api_key_here

---

Step 3 – Set Up MySQL Database

1. Start XAMPP → launch MySQL
2. Open phpMyAdmin
3. Create a database named ai_chat
4. Run the following SQL script:

CREATE TABLE conversations (
    id INT AUTO_INCREMENT PRIMARY KEY,
    user_message TEXT,
    ai_response TEXT,
    timestamp TIMESTAMP DEFAULT CURRENT_TIMESTAMP
);

---

Step 4 – Run the Application

streamlit run ui_app.py

Then open http://localhost:8501 in your browser.

---

💬 Application Modes

🗨️ Chat Mode

• Open the Chat tab
• Type a message → click Send
• Gemini 2.0 Flash replies instantly
• Each conversation is saved automatically in MySQL

📄 RAG Mode

• Open the RAG tab

• Upload a .txt file → click Start Conversation

• Ask questions directly about the uploaded document

• The system retrieves the most relevant chunks using cosine similarity

• Gemini generates context-aware answers

• You can:

  • View highlighted relevant text
  • Check similarity scores
  • Search specific keywords
  • Export chat history
  • Reload the uploaded file

---

🔍 How RAG Works

1. The uploaded text file is split into overlapping chunks

2. Each chunk is embedded using a Sentence Transformer

3. When a user asks a question:

   • The app finds top-matching chunks via cosine similarity
   • These chunks are given to Gemini 2.0 Flash as context

4. Gemini responds using only the retrieved information

5. The UI shows answers with highlighted contexts and relevance scores

---

📸 Screenshots

💬 Chat Interface

📄 RAG Interface – Question 1

📄 RAG Interface – Question 2

📁 Ensure screenshots are located in a folder named ss in the project root:

project_folder/
├── ui_app.py
├── rag_engine.py
├── chat_logic.py
├── db_config.py
└── ss/
    ├── chat_interface.png
    ├── rag_interface_1.png
    └── rag_interface_2.png

---

✅ Deliverables Checklist

Item	Status
Source Code (GitHub / ZIP)	✔️
Short README File	✔️
Working AI Chat + RAG App	✔️
Screenshots Included	✔️

---

👨‍💻 Developer Information

• Name: Dimuthu Shalinda
• University: University of Jaffna – Faculty of Engineering
• Project: AI Chat Application with RAG
• Date: October 2025

---

🏁 Summary

This project highlights:

• Real-time chat with Google Gemini 2.0 Flash

• A full Retrieval-Augmented Generation pipeline

• Persistent conversation logging via MySQL

• Advanced features:

  • Highlighted context visualization
  • Similarity scoring
  • Keyword search capability
  • Chat export and file reload options

This system combines LLM reasoning with retrieval-based knowledge grounding, demonstrating how modern AI can deliver accurate, context-aware responses from custom documents.