Shelf Vision — AI Retail Audit
LiveFull-stack build · AI integration
Upload a photograph of a retail shelf and the platform returns a structured product inventory: product name with brand and size, category, visible facings count, and a per-item confidence level — all in Arabic.
The AI work is the substance here. A Gemini vision model runs behind a Zod response schema, so the model returns typed, validated objects rather than prose I would have to parse. The prompt is written as a retail-shelf-auditing brief with an explicit empty-array contract for non-shelf images, and the route carries a 60-second budget because a dense shelf photo genuinely needs it. Every field is described in the schema so the model knows what each one means.
p.onemedia24.xyz- Front-endNext.js 16, React 19, TypeScript
- AIVercel AI SDK, Gemini vision, Zod schema
- TypeStructured output, typed generation
- VerifiedHTTP 200 · 1.9 s