TasteShip

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for TasteShip

The problem

Launch copy without cultural grounding sounds the same for every audience. Generic LLMs invent vibe; they do not know what people actually like across film, music, dining, fashion, and brands.

The solution

TasteShip turns a pasted feature ask into a same-day ship pack using an agentic server-side tool loop that calls Qloo:

1. parse_ask — audience + product intent
2. qloo_search — GET /search on https://hackathon.api.qloo.com
3. select_signals — pick interest entities
4. qloo_insights — GET /v2/insights with signal.interests.entities
5. compose_ship_pack — scope, build vs fake-door, changelog, launch tagline, customer replies, and a taste rationale that names Qloo entities

The UI shows the tool-call trace so judges can see Qloo in the loop. Launch copy may name an entity only if Search/Insights returned it; with the live key set, empty taste results leave those sections blank (no mock fill-in).

Built with

• Qloo Taste AI Hackathon API (Search + Insights)
• Next.js App Router, TypeScript, Tailwind
• Server-side agentic tool loop

Demo for judges

Open https://tasteship.vercel.app — paste a feature ask or use a sample path. /api/health reports live mode when QLOO_API_KEY is set.

Repo: https://github.com/icohangar-ops/tasteship (MIT)

Built With

• agentic-ai
• next.js
• qloo
• typescript