Qloo Agentic Hackathon
Cast a weekend from what you already love.
Name a film, a record, a book, or a mood. TasteCast looks those names up in Qloo, then asks the taste graph for the things that sit beside them: another film, the artists, the table, the city.
Three ways in
Each one opens the desk with the line already written.
I love Studio Ghibli and rainy-day jazz. Cast me a weekend in Brooklyn: a film, playlist artists, and neighborhood dining.
Cross-domainOne title, other roomsSpirited Away
Books and recordsDarker shelfI love Radiohead and Neverwhere. Cast books, artists, and a film, and keep it dark.
How Qloo powers this
A language model can plan the calls. It cannot invent the board. Entity ids come from search, tag ids come from the tag index, and every recommendation is an entity Qloo returned for those signals. That is why a Ghibli seed can surface a Brooklyn restaurant or a book you would not get from a genre keyword alone: the link lives in the taste graph.
With no OPENAI_API_KEY, a rule-based planner still runs the same Qloo calls. Set the key when you want the planner to read looser sentences.