VELARI

Services

• Web Application
• REST API & Backend
• AI-Powered Features
• UI/UX Design

Stack

Overview

Velari is a B2B platform for private-label cosmetics brands. Companies create an account, manage their brand catalogue, attach cosmetic formulas to products, and place orders — all within a multi-tenant, role-based system. The platform is available in three languages and includes a full AI layer for product content generation and a semantic knowledge base for the chat assistant.

Challenge

Private-label cosmetics involves complex relationships between companies, brands, formulas, and orders. Each company needs isolated data, per-brand product catalogues, flexible ordering rules, and a team of people with different roles — all while keeping the interface simple enough for non-technical users in three different locales.

Solution

Built a Django REST API with a five-role RBAC system (Superuser, Staff, Salesman, Company Owner, Company Member), MPTT hierarchical categories, and per-company minimum order quantities. Authentication uses HttpOnly JWT cookies to eliminate XSS risks. The Next.js frontend supports three locales via next-intl. AI is integrated at multiple points: Claude and GPT models generate product names, INCI ingredient lists, and colour descriptions on demand. A pgvector-backed knowledge base auto-embeds documents on upload and enables semantic search in the chat assistant. Invoices can be generated directly from chat via JSON parsing of structured AI responses.

Outcome

A production-ready B2B platform deployed on DigitalOcean (backend) and Vercel (frontend), with Nginx, SSL/HTTPS, and HSTS. The AI model selection API allows switching between Claude Haiku/Sonnet/Opus and GPT-4o/4o-mini at runtime without code changes.