Task

PDFImageMerger addresses a practical need: PDFImageMerger is a Python tool designed to merge images into a single PDF file, specifically for folders named 'Passport'. It supports various image formats, automatically detects image orientation, and excludes system and hidden folders from processing.

The project is part of commercial development and is completed and in use. It is related to services: Full-stack web applications and dashboards.

The case demonstrates an approach to tasks in the area of "Full-stack web applications and dashboards": from requirements to delivery with a clear module structure.

Before starting, readiness criteria were established: a clear first screen, predictable form errors, and the possibility of modifications without rewriting half of the codebase.

Solution

A closed commercial tool has been implemented: the interface and logic are tailored to everyday user scenarios, without unnecessary screens.

The scenarios are designed so that the main actions fit within one or two screens: fewer switches, fewer input errors. Navigation and field labels are aligned with familiar patterns—users do not need to "learn the interface".

Special attention is given to loading states and empty lists: skeleton/spinner where data arrives asynchronously, and clear empty-state texts instead of an empty table.

Architecture and stack

Infrastructure and utilities: TeX.

Language distribution in the repository: TeX (52.9%), HTML (46.6%), Python (0.4%).

Layers are separated so that the UI does not know the details of transport: API/storage adapters, domain types, presentation components. This simplifies backend replacement and targeted UX adjustments.

Key features

• TeX — used in product logic: reduces the amount of custom code and speeds up the addition of new screens.

Additionally:

• Integration and use of TeX in product logic and interface.
• Validation of user input and clear error messages.
• Responsive layout for desktop and mobile.

Result

The project has been brought to a working state: the team/user receives a stable tool instead of manual tables and correspondence. Time for routine operations has been reduced due to automation in the interface.

The code is closed per client requirements; architectural solutions and stack are described on this page for transparency of expertise and SEO.

What I would improve

The next step is observability (logs/metrics), regression auto-tests, and API documentation for integrators. For high-load scenarios—profiling bottlenecks and caching reference data.