AI APP · IN DEV · CASE STUDY

Scan a QR. Ask the product a question.

QR codes linked to product guide pages with an AI assistant strictly grounded in the product manual — no hallucinations, no off-topic answers.

Every physical product has a manual. Most customers never read it. Product QR Guides turns the manual into a conversation — the customer scans a QR code, opens a hosted guide page, and asks the product anything. The AI answers from the manual and only from the manual.

Build something like this
WHAT WE BUILT

One QR code per SKU. One AI per manual.

Physical-to-Digital
Unique QR per Product

Each physical product or SKU gets its own unique QR code that links to a hosted product guide page. The page displays key specs, navigable manual sections, and the AI chat widget — all branded per client.

Grounded AI
Claude Sonnet, Manual-Only Answers

The AI chat assistant is powered by Claude Sonnet and strictly grounded via system prompt in the product's manual content only. The model is instructed to refuse any question outside the manual scope — hallucination risk is eliminated by design, not by luck.

Ingestion Pipeline
PDF → Chunks → Vectors

Manual ingestion pipeline: PDF or text manuals uploaded to Supabase, chunked and embedded for retrieval-augmented generation. At query time, the most relevant chunks are retrieved and injected as context — the AI answers from actual manual text, not from parametric memory.

Multi-Tenant
Brand and SKU Isolation

Each brand and SKU gets fully isolated manual content and isolated chat history. One deployment serves any number of clients and products — a refrigerator brand and a power tool brand share no data and no context.

AI & RETRIEVAL

How the grounded RAG works.

STEP 01
MANUAL INGESTION

PDF or text manual is uploaded, parsed, chunked into semantically coherent sections, and embedded using an embedding model. Vectors and source text are stored in Supabase with product and brand identifiers attached to every row.

STEP 02
RETRIEVAL

Customer question is embedded and matched against the product's manual vectors using cosine similarity. Top-k most relevant chunks are retrieved from that product's isolated namespace only — no cross-product context leakage.

STEP 03
GROUNDED GENERATION

Retrieved chunks plus the customer question are passed to Claude Sonnet with a system prompt that explicitly instructs: answer only from the provided manual excerpts, refuse any question outside scope. The AI cannot answer questions the manual doesn't cover.

STACK

What it runs on.

Claude Sonnet
Grounded RAG chat
Next.js
Guide page + chat UI
Supabase
Vector embeddings + storage
Vercel
Hosting + edge functions
pgvector
Similarity search
PDF parser
Manual ingestion
IN DEVELOPMENT

Every product deserves an AI
that actually knows it.

Grounded AI chat for your product line. Initial client deployment in progress — get in touch to join.

Start the conversation →