Overview

This is an MSME (Micro, Small & Medium Enterprise) Business Management Platform built with Streamlit. The application provides role-based access control for Admins and Shop Owners, featuring inventory management capabilities, document-based knowledge retrieval using RAG (Retrieval-Augmented Generation), and AI-powered chat assistance. The platform enables businesses to manage their inventory through CRUD operations, upload and query documents for knowledge management, and interact with AI chatbots for business assistance.

System Architecture

Frontend Architecture

• Framework: Streamlit web application framework
• Layout: Wide layout with expandable sidebar for navigation
• Authentication: Token-based authentication with role-based access control
• Session Management: Streamlit session state for user authentication and role persistence

Backend Architecture

• Application Structure: Modular design with separate utility modules for different functionalities
• Core Modules:
  • auth_utils.py: Authentication and authorization logic
  • inventory_utils.py: Inventory management operations
  • rag_utils.py: Document processing and knowledge retrieval
  • chat_utils.py: AI chat functionality
  • database_init.py: Database initialization and schema setup

Data Storage Solutions

• Primary Database: SQLite for inventory management
  • Single table design with auto-incrementing primary keys
  • Indexed product names for optimized search performance
  • Timestamp tracking for creation and updates
• Vector Storage: FAISS (Facebook AI Similarity Search) for document embeddings
  • Separate vector stores for Admin and Shop Owner roles
  • Persistent storage using pickle for text chunks
  • Flat L2 distance indexing for similarity search

Authentication and Authorization

• Token-based Authentication: Simple token validation system
• Role-based Access Control: Two distinct roles (Admin and Shop Owner)
• Configuration: Environment variables and Streamlit secrets for token management
• Session Persistence: Authentication state maintained in Streamlit session

AI and Machine Learning Components

• Embedding Model: SentenceTransformers 'all-MiniLM-L6-v2' for text vectorization
• Language Model: Google Gemini 2.5 Flash for response generation
• RAG Implementation: Custom retrieval-augmented generation for document queries
• Document Processing: PDF text extraction using PyPDF2

Data Processing Pipeline

• Inventory Management: CSV import/export functionality with schema validation
• Document Processing: PDF text extraction and chunk-based indexing
• Vector Search: Similarity-based document retrieval for contextual responses

External Dependencies

AI Services

• Google Gemini API: Primary language model for chat responses and business assistance
• SentenceTransformers: Hugging Face model for text embeddings and similarity search

Data Processing Libraries

• pandas: Data manipulation and CSV operations for inventory management
• PyPDF2: PDF document parsing and text extraction
• FAISS: Vector similarity search and indexing
• numpy: Numerical operations for vector computations

Database and Storage

• SQLite: Embedded database for inventory data persistence
• pickle: Python object serialization for vector store text storage

Web Framework

• Streamlit: Complete web application framework with built-in session management and secrets handling

Configuration Management

• Environment Variables: Token configuration and API key management
• Streamlit Secrets: Secure configuration storage for production deployment