Our Technology
The technical infrastructure behind automated food compliance.
The 4-Layer Architecture
Each layer is modular, testable and independently scalable.
Document Pipeline
Multi-format ingestion: PDF, scan, photo, email. AI-powered extraction with confidence scoring and human-in-the-loop.
Multi-provider AI (Claude 3.5, GPT-4o, Qwen-VL) with confidence scoring. Supports menu photos, PDF and label scans. Fallback chain on low confidence.
Product Graph
EAN-verified knowledge base with 27,000+ nodes. Label data as canonical source. Automatic deduplication.
PostgreSQL-based product graph with 27,000+ EAN-verified nodes. Automatic label verification per EU Regulation 1169/2011. Deduplication via manufacturer+packaging hash.
Compliance Engine
Rule-based allergen model per EU Regulation 1169/2011. Hierarchical cascade: Ingredient → Component → Dish. Auditable.
Rule-based cascade: Ingredient → Component → Dish. Consistency guarantee via trigger-based propagation. Audit trail for every allergen calculation.
Publishing Layer
Multi-tenant WordPress Multisite with i18n (DE/EN/ZH), Schema.org JSON-LD and AI-generated description per dish.
WordPress Multisite with dedicated BLOG_ID per customer. i18n pipeline (DE/EN/ZH) via AI translation. Schema.org JSON-LD with 9-entity graph.
Tech Stack
AI Pipeline
Multi-provider routing: Claude (Anthropic), GPT-4 (OpenAI), Qwen-VL (Alibaba). Regional privacy via configuration.
PostgreSQL + EAN Graph
Relational knowledge base with 50+ tables, CHECK constraints, materialized views for allergen cascade.
FastAPI + REST
80+ endpoints. JWT auth, multi-tenant scoping, rate limiting. Uvicorn-based on Hetzner VPS.
WordPress Multi-Tenant
Multisite with custom domains (CNAME). 19 shortcodes, customer-specific dashboard, role-based access.
© 2026 GastroLogic System