TRUKBK
I engineered and launched the digital storefront for TRUKBK, a premium UK-based manufacturer of modular aluminium service bodies and truck trays. The goal was to create a high-end web experience that matched the rugged, premium nature of their physical products, while automating their sales pipeline using advanced AI. Selling custom, £10k+ physical truck builds is a high-touch process. Relying on a generic AI chatbot was dangerous—if an AI hallucinates a physical specification or price, it could lead to costly returns and broken trust. TRUKBK needed an automated way to answer customer queries with 100% factual accuracy and capture hot leads 24/7.
View Live Site
Core Features
Industrial Noir Aesthetic
Implemented a dark, premium UI utilizing deep blacks, stark whites, and industrial orange accents with glassmorphism and micro-animations.
Interactive Configurator
Developed a dynamic, 6-stage product builder that allows users to select their vehicle, cab type, body style, finish, and accessories.
Zero-Hallucination AI
Custom Retrieval-Augmented Generation (RAG) pipeline using Google Gemini 1.5 Flash and Supabase (pgvector).
Lead Generation
Integrated Web3Forms for seamless, serverless contact form submissions.
Technical Deep Dive
Retrieval-Augmented Generation (RAG) Architecture
Knowledge Base Ingestion: All proprietary product specs, pricing tiers, and vehicle compatibility charts were chunked into precise logical segments and converted into high-dimensional vector embeddings stored in a Supabase PostgreSQL database using pgvector. During live semantic search, queries are embedded in real-time and matched via cosine similarity. The verified facts are then injected into the Gemini prompt to completely eliminate hallucinations.
The 2-Layer Bulletproof Fallback System
APIs fail, and quotas run out. To ensure the business never looks broken to a customer, I engineered a robust 2-layer fallback system. Layer 1 is the full Gemini AI RAG pipeline. Layer 2 is a custom Rule-Based Engine that silently takes over if the API times out. It uses keyword matching against a hardcoded knowledge base to return accurate prices and contact info without the user ever seeing an error message.
Performance Benchmark
"Guaranteed 100% uptime for the chat experience via a custom fallback architecture, ensuring no leads are lost to technical errors while reducing the manual sales support load."