Component and Technology Overview

This document provides an overview of the key components in the system and the technologies used.

Frontend Application

Technology: React with React Router (JavaScript)

The frontend is built using React, offering a component-based architecture for interactive user interfaces. React Router enables seamless navigation within the application.

Key Features:

• Component-Based Design: Modular and reusable components for maintainability and scalability.
• State Management: Utilizes React Hooks like useState, useEffect, for efficient state handling.
• Dynamic Routing: React Router enables smooth navigation between pages.

Server

Technology: Flask (Python)

The backend processes user queries, retrieves relevant data, adds new documents, and communicates with the LLM.

Key Features:

• Query Processing: Parses and validates user queries before searching for relevant documents.
• Document Retrieval: Interacts with the Indexed Documents Database to fetch relevant information.
• Response Handling: Constructs a prompt for the LLM and returns the generated answer to the frontend.
• Error Handling & Logging: Implements monitoring, logging, and error handling for reliability.
• Admin Controls: Allows administrators to add new documents to the database, ensuring the system stays up to date.

SQL Database and Vector Database

Technology Stack: PostgreSQL with PGVector

The system utilizes PostgreSQL for structured data management and PGVector for efficient vector-based searches.

Database Functions:

• SQL Database:
  • Stores admin user credentials for secure backend dashboard access.
  • Manages metadata and binary data of uploaded documents.
• Vector Database:
  • Stores vectorized chunks of uploaded documents for similarity search.
  • Ensures efficient indexing and retrieval of relevant content.

Key Features:

• Full-Text Search:
  • Enables rapid retrieval of relevant documents based on query matching.
• Document Updates:

  • Supports continuous addition and deletion of documents.

Large Language Model (LLM)

Technology: Llama

The LLM generates answers by processing the prompt constructed using retrieved documents.

Key Features:

• Context-Aware Responses: Uses relevant documents to improve the accuracy of answers.
• API-Based Access: Communicates with the LLM hosted locally via API calls.

How Technologies Support System Requirements

This stack ensures a scalable, efficient, and high-performance system for answering user queries.

• React + Flask + PostgreSQL: React enables an interactive UI, while Flask handles backend queries and PostgreSQL efficiently stores and retrieves documents.
• Vector Database + LLM: PGVector stores vectorized documents for fast retrieval, and DeepSeek/Llama enhances response accuracy by incorporating document context.

This stack combines performance, flexibility, and cost-effectiveness for a responsive, evolving system.