Architecture & Design

This section provides a comprehensive overview of the system architecture, application structure, and key design decisions for the Fluidsdata Digitization and OCR application.

Architecture Contents:

• App Structure
  • Overview
  • Root Structure
  • Core Application Components
  • Upload/Processing Data Flow
  • Technology Integration
  • Performance Considerations
  • Security & Data Handling
• Upload Pipeline
  • File Handling
  • OCR Process
  • Table Reconstruction

Overview

1. Frontend Layer (App Structure):

   • Streamlit multi-page application

   • Session management for state persistence

   • Real-time processing feedback

2. Processing Layer (System Architecture):

   • Document processing pipeline

   • OCR and table reconstruction using DocTR and YOLO models

   • Data mapping and merging capabilities

3. Backend Layer (Azure Backend):

   • Custom Azure-based API for data handling

   • Integration with Azure Blob Storage for file management

   • Database interactions for structured data storage

4. State Management (Session Management):

   • AppSession for processing workflows

   • DigitizationSession for analysis workflows

   • Persistent state across user interactions

Design Principles (somewhat adhered to)

• Modularity: Each component is designed to be independent and reusable.

• Separation of Concerns: Clear distinction between UI, processing, and data layers.

• Industry Focus: Tailored for oil & gas PVT report digitization.

Note

The existence of two state management classes (AppSession and DigitizationSession) is a relic of earlier development where two separate apps (OCR and Digitization) were merged together. This should be refactored in the future to simplify state management and improve code clarity.