Menu

Loading

GzoneAI

The
Challenge

GzoneAI is a full-stack project and material management platform built for the manufacturing and construction industry. The system handles the full lifecycle of production projects — from CSV-based work order imports and AI-powered material extraction to customer management, part tracking, costing, and detailed reporting — all in one place.

Managing manufacturing or construction projects means dealing with large, complex material lists buried inside raw project descriptions and CSV exports. Before GzoneAI, teams faced:

  • Manual interpretation of unstructured project descriptions to identify parts, dimensions, and quantities — slow and error-prone
  • No centralised system linking customers, projects, materials, and costs in one view
  • Repeated re-upload of identical files with no deduplication, causing data drift
  • Material pricing scattered across spreadsheets with no version control or multiplier management
  • No automated notifications when file processing or AI jobs succeeded or failed

The
Solution

A purpose-built, full-stack web application that serves as the single source of truth for every project — from initial file upload through AI-driven part extraction to final material reporting. An OpenAI Assistant processes raw project descriptions automatically, eliminating the need for manual data entry of dimensions and quantities.
Project & Customer Management
  • Full CRUD for projects: code, name, description, assigned manager, registered/assigned/sent/deadline dates
  • Customer profiles with type classification and contact metadata
  • Project expense tracking alongside lifecycle dates
CSV File Upload & Processing
  • Semicolon-delimited CSV uploads with automatic parsing and line-level storage
  • MD5-based file deduplication — re-uploading an identical file is detected and rejected
  • Uploaded lines auto-create elements and parts in the database
  • 50 MB upload limit with full error handling
AI-Powered Material Extraction
  • OpenAI Assistant integration parses unstructured project descriptions
  • Extracts part names, dimensions (x/y/z), quantity, and thickness automatically
  • Calculates derived fields (area, perimeter) from extracted dimensions
  • Results feed directly into part and element records — no manual re-entry
  • Email notification on AI job success or failure
Material Catalog & Pricing
  • Material catalog with hierarchical categories
  • Per-material fields: code, vendor/external code, unit price, unit type, currency (default HUF)
  • Configurable price multipliers and sale price adjustments per material
  • Many-to-many part–material associations
Reporting
  • Aggregated material report: summary of all materials required across a project
  • Detailed material report: breakdown by element and part
  • Project file browser for all associated documents
User & Access Management
  • JWT (RS256) + PASETO v4 authentication with Argon2 password hashing
  • Session expiry: 12 hours default, 14 days with "remember me", 90 days for root
  • Role-based access control via CASL (root, admin, and configurable roles)
  • API key authentication for programmatic/external access
  • Swagger/OpenAPI documentation at /system/doc
File Storage
  • Local temp and project file directories for development
  • AWS S3 as primary file store in production
  • AWS Secrets Manager and Parameter Store for credential and config management

Scope &
Scale

  • Full project lifecycle management — from CSV import through AI extraction to final cost reporting
  • OpenAI Assistant integration with automatic element/part creation from unstructured descriptions
  • Dual authentication (session JWT + API key) with role-based CASL authorization
  • AWS-native production stack — S3, Secrets Manager, Parameter Store
  • Two frontend variants (Vite SPA + Next.js 14) sharing the same backend API
  • Hungarian-first localization across all UI components and date formatting

The
Outcome

GzoneAI replaced manual CSV wrangling and spreadsheet-based material lists with a single, integrated platform.

Uploading a project file now triggers automatic parsing, deduplication, and AI-powered dimension extraction — turning an hour of manual data entry into a background job that completes in seconds. Material reports that previously required consolidating multiple spreadsheets are now generated on demand, always reflecting the current state of the project. Pricing updates propagate instantly across all parts and reports, and email alerts mean the team is notified the moment an AI job finishes — or fails.