Inspiration
Independent venues often have little purchase history for a first-time buyer. SecondShow explores how a venue marketer can recommend a relevant next show from the venue's own upcoming lineup while keeping eligibility and human review in the workflow.
What it does
SecondShow builds a reviewable campaign for eligible first-time buyers using fictional sample data or anonymized CSV imports. It freezes a cohort and filters future events by consent, purchase history, inventory, date, budget and admission constraints. The operator can compare a transparent genre/date baseline with Qloo-assisted planning, review recommendations and editable copy, approve the campaign, and export CSV or JSON.
The prototype also includes a fictional recipient preview, simulated checkout and a 30-day outcome view. It does not send messages, issue tickets or process payments.
How we built it
The Next.js and TypeScript application keeps workspace data in browser-local storage. In live planning, the server asks Qloo to rank confirmed artist identities from a feasible shortlist. A bounded tool-calling workflow then uses NVIDIA Nemotron 3 Super (nvidia/nemotron-3-super-120b-a12b) through the Nebius Token Factory API to inspect candidates and propose events using temporary aliases. The application validates the returned event IDs and requires human approval before export. Provider errors remain visible; the operator must explicitly choose a baseline fallback.
What we tested
A real nonpersonal test returned successful responses from Qloo and two Nebius completions for one fictional candidate. A separate local browser flow generated a live proposal, required human approval and enabled export. These checks demonstrate the integration and review flow; a one-candidate test does not establish comparative ranking quality, conversion lift or production readiness.
What's next
The public judge demo runs the fictional sample workflow with live provider calls disabled. The video shows a separate recorded live integration test. A future evaluation would compare the same feasible catalog ordered by the baseline and by Qloo-assisted planning, with a predeclared repeat-purchase endpoint. All sample bookings, buyers, inventory and purchases are fictional.
Built With
- nebius-token-factory
- next.js
- nvidia-nemotron
- playwright
- qloo
- react
- typescript
Log in or sign up for Devpost to join the conversation.