QR-Based Web App with AI Chat Widget — Multi-language, Real-time Data Sync

We're building a multi-tenant web platform where each client gets their own branded, QR-code-accessible page displaying a dynamic content catalog (items with name, description, price, category, and metadata such as tags/attributes). Each page includes an embedded AI chat assistant that answers visitor questions grounded in that specific client's live catalog data — recommendations, attribute lookups, translations, and general Q&A strictly scoped to that content. Multi-tenant architecture — each client has their own account/login, isolated data, and a unique URL/QR code. No shared state between clients. Content management dashboard — clients can add/edit/delete catalog items (name, price, description, attributes, tags, category) through a simple admin UI. AI chat assistant — embedded chat widget grounded on the current state of that client's catalog (no retraining step; live data is passed as context on every request). Must be scoped to only answer questions relevant to that client's content and politely decline unrelated topics. On-demand translation — end users can ask the chat assistant, in natural language and in any language, to display the catalog in a different language; the catalog view updates accordingly and caches the translation. Per-client branding — each client can upload a logo, pick a brand color, and select a typography style; these apply dynamically across their catalog page and chat widget via a theming system (CSS variables / design tokens), with automatic accent-color suggestion extracted from the uploaded logo. Payment/subscription billing — Stripe integration for recurring monthly billing per client account. Usage analytics — basic dashboard showing client-level metrics: page views, chat questions asked (aggregated/anonymized), most-viewed items. Backend proxy for LLM API calls — all AI calls must go through our backend (not client-side), with rate limiting and cost controls to prevent abuse. GDPR-compliant data handling — we are EU-based; personal/usage data handling must comply with GDPR from day one.