MosaicHour

Inspiration

Life enrichment staff plan activities for residents with different histories, interests and access needs. One activity director described planning two or three events a day for about 50 residents on a $300 monthly budget, while repeated bingo and bowling had lost their appeal. NCAL guidance encourages activities that reflect residents' interests and abilities. MosaicHour gives senior living teams a starting point grounded in the group's own cultural favorites.

What it does

Enter a musician and film that seniors in the group enjoy, along with group size, session length, supply budget and an access priority. MosaicHour returns three activity drafts: a music gathering, a film discussion and a story circle. Each includes a Qloo-related cultural anchor, facilitation steps and a staff review note. Staff choose an idea and print a run sheet.

Try the prefilled Frank Sinatra and Casablanca example. In a live run, Qloo connected the favorites to artists such as Nat King Cole and Billie Holiday, films such as It's a Wonderful Life, and books such as The Great Gatsby. The app shows the candidate recommendations and agent trail so staff can see what informed the ideas. The AI-written activity descriptions are drafts; staff decide what fits their group.

How it uses Qloo and agents

Each plan makes five Qloo API calls: two searches resolve the musician and film, then three Insights calls find related artists, films and books using both favorites together. These Qloo results supply the cultural anchors; Gemini turns them into activity drafts. A LangGraph workflow coordinates four steps: Qloo Scout, Taste Bridge, Program Designer and Access Reviewer. The reviewer checks that selected anchors came from Qloo results and adds staff review notes. A logged template fallback keeps the workflow usable if Gemini is unavailable.

FastAPI serves the app on Google Cloud Run. Staff selection is recorded by the backend and enables a printable run sheet. The app does not schedule or send activities. It does not ask for resident names, diagnoses or medications.

Challenges and what's next

A language model can suggest an entity that was not returned by Qloo, so the workflow checks its chosen anchors against the actual results. Activities also need human review for accessibility, suitability, availability and content rights. We have not yet tested MosaicHour with a senior living team or measured planning time or participation outcomes. Next, we want to interview activity coordinators and learn whether the recommendations help them plan sessions residents want to join.

Built with

Qloo API, Gemini, LangGraph, FastAPI, Python, HTML, CSS, JavaScript and Google Cloud Run.

Built With

• qloo