ShelfOps

ShelfOps is a controlled Next.js hackathon demo that converts one retail shelf photo into live operational tasks. It compares real Cerebras Gemma 4 inference against local Gemma inference through Ollama to show why low-latency AI matters for retail shelf auditing.

Demo claim

Cerebras makes shelf auditing usable in the aisle. Local inference makes it a delayed report.

Highlights

• Retail shelf audit from a single shelf photo
• Cerebras Gemma 4 31B provider path
• Local Gemma on Mac through Ollama
• Sequential agent workflow with real timings
• Planogram, pricing, and operations task generation
• One-viewport hackathon demo built with Next.js App Router and TypeScript

Tech stack

• Next.js App Router
• TypeScript
• Tailwind CSS
• Zod
• Cerebras Chat Completions API
• Ollama local API
• Native Server-Sent Events

Run locally

npm install
cp .env.local.example .env.local
npm run dev

Open http://localhost:3000.

Environment

CEREBRAS_API_KEY=your_key
CEREBRAS_BASE_URL=https://api.cerebras.ai/v1
CEREBRAS_MODEL=gemma-4-31b
OLLAMA_BASE_URL=http://localhost:11434/api
OLLAMA_MODEL=gemma4:12b-mlx

Validate

npm run lint
npm run build