📊 Segment Tree-Enhanced Retrieval in AI Pipelines

Case Study for Data Structures and Algorithms (23CSE203)

This project integrates Data Structures (Segment Tree) with an AI-powered Retrieval-Augmented Generation (RAG) pipeline to optimize stock-related queries. It demonstrates how advanced algorithms and AI models can work together for efficient data retrieval, analysis, and user interaction.

---

🚀 Project Overview

• Backend (Python + Flask)

  • Implements Segment Trees for efficient stock query operations (min, max, sum, average over a date range).
  • Integrates with Pinecone Vector DB to store embeddings.
  • Uses Google Gemini embeddings + LLM for natural language queries.
  • Provides APIs that combine AI + data structure outputs.

• Frontend (React + Vite)

  • Interactive UI for querying stock data.
  • Displays AI-enhanced insights along with structured results from the segment tree.
  • User-friendly interface to test queries.

---

Work Flow

Segment Tree + RAG Pipeline Flow

User Query
↓
Flask API (/chat endpoint)
↓
chat_bot.py (Query Handler)
↓
RAG Pipeline (rag_setup.py)
|   ├─ Pinecone Vector DB (document retrieval)
|   ├─ Google Gemini Embeddings (text → vectors)
|   └─ Gemini LLM (natural language understanding)
|                      ↓
|       Generates JSON Command OR Text Answer
↓
Decision Point:
    ├─ [JSON Command Found] → Segment Tree (segment_tree.py)
    │                    ↓
    │            Numerical Result
    └─ [No Command] → Direct RAG Answer
                         ↓
                    Final Answer

Description

---

Description

• User Query: Natural language input from the frontend UI.
• Flask API: Receives user query and routes to backend query handler.
• chat_bot.py: Handles preprocessing and determines if Segment Tree is needed.
• RAG Pipeline:
  • Retrieves relevant documents using Pinecone Vector DB.
  • Converts text to embeddings using Google Gemini.
  • Uses Gemini LLM for understanding and generating a JSON command or direct text answer.
• Decision Point:
  • If a JSON command is generated (e.g., range query for min/max/sum/avg), it is processed by the Segment Tree module for efficient computation.
  • If no JSON command is generated, the response is taken directly from the RAG output.
• Segment Tree (segment_tree.py): Performs efficient numerical operations for the query.
• Final Response: Sent back to frontend for display.

---

📂 Project Structure

├── backend/ 
├── frontend/ # React + Vite app
├── .gitignore
├── README.md
└── LICENSE   

---

⚙️ Installation & Setup

🔹 Backend Setup (Python)

# Navigate to backend
cd backend

# Create a virtual environment
python -m venv venv

# Activate the environment
# On Windows
venv\Scripts\activate
# On Mac/Linux
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Run backend server
python app.py

The backend will start running (default: http://127.0.0.1:5000/).

🔹 Frontend Setup (React + Vite)

# Navigate to frontend
cd frontend

# Install dependencies
npm install

# Start development server
npm run dev

The frontend will run on http://localhost:5173/ by default.

---

🧠 How It Works

User Query → User enters a stock-related query in natural language (e.g., "What is the average price of AAPL between Jan 1 and Jan 15?").

RAG Retrieval → Gemini embeddings + Pinecone fetch relevant historical stock data.

Segment Tree Processing → If the query involves numerical operations (min, max, avg, sum), a Segment Tree processes it efficiently.

AI + Data Structure Response → Final response is generated by combining retrieved AI context and Segment Tree calculations.

Frontend Display → Results are displayed neatly in the React UI.

---

✨ Features

• 🔍 Hybrid Querying → Combines AI retrieval with Segment Tree efficiency.
• 📈 Efficient Stock Queries → Range queries (min, max, avg, sum) in O(log n).
• 🤖 AI-Powered Assistant → Understands user queries in natural language.
• 🌐 Full-Stack Integration → Backend (Python) + Frontend (React + Vite).