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.

.. toctree::
    :maxdepth: 2
    :caption: Architecture Contents:

    app_structure
    upload_pipeline/index

Overview
-----------------

1. **Frontend Layer** (:doc:`app_structure`): 
    * Streamlit multi-page application
    * Session management for state persistence
    * Real-time processing feedback

2. **Processing Layer** (:doc:`system_architecture`): 
    * Document processing pipeline
    * OCR and table reconstruction using DocTR and YOLO models
    * Data mapping and merging capabilities

3. **Backend Layer** (:doc:`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** (:doc:`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.