Skip to main content

ShelfScout

Restocking reccomendator for pet stores.

About the project

Which pet products should your store stock? Paste an online pet store’s link. ShelfScout reads the store, then ranks 8 pet-supply niches for it using ~40,000 real Amazon reviews, each labelled by a System One decision model. Every number links back to the reviews behind it.

Amazon gives sellers niche-level review insights, but only for selling on Amazon. ShelfScout gives independent stores (Shopify and others) a fit verdict for their store, built on counted evidence instead of a chatbot’s opinion.

How it works

  • Offline: Amazon Reviews 2023 (McAuley Lab), Pet Supplies → 8 niches · 205 products · 39,996 reviews in SQLite. Each niche has 6–7 aspects (e.g. dog beds: support, durability, washability). Drex Max labels every review × aspect as praised / complained / not mentioned, with a confidence score. Labels under 0.6 confidence are left out of the counts.
  • Live (~11 s, ~$0.001 per store):
    1. Gemini profiles the store: animals, price band, audience, feel.
    2. For each niche, Drex weighs each aspect for this store and returns a typed verdict: fit probability, which product to start with, risk 0–4, and the deciding aspect.
    3. SQL builds a four-box: table stakes · differentiators · category gaps · red flags.
    4. Gemini writes an honest overview using only those numbers and real review quotes.

Stack

Next.js 16 · React 19 · Tailwind 4 · shadcn/ui · SQLite · Python data pipeline (stdlib only) · OpenRouter (Drex Max via the Decisions API + Gemini 3.5 Flash Lite, one key)

Next

Pull fresh Petco reviews through Bright Data for a live “2023 vs today” comparison.

Built with

elevenlabsopenrouterhuggingface

AI tools

Coded withClaude Code

Prize Categories Entered

  • Gemini Prize Category Q1 2025 - Q1 2026

    Best Use of Gemini API

    Submitted

    Gemini model for web search.

  • HyperFrames

    Best Demo Video Using HyperFrames

    Submitted

    Use this to build demo video from scratch.