Dayger

Inspiration

Mornings are chaos. You're tired, indecisive, and somehow expected to pick an outfit, find food, and mentally prepare for a midterm — all before 9am. We wanted to kill that gray area entirely.

What it does

Dayger takes a single morning prompt and returns five coherent, personalized cards: real-time weather, a weather-matched outfit, a mood-tuned playlist, an energy/focus plan, and UCLA meal picks — all consistent with each other and ready in seconds.

How we built it

React + Vite frontend with progressive card loading, a Node/Express/TypeScript backend orchestrating a multi-agent Python pipeline, PostgreSQL for anonymous-first user persistence, and agents powered by ASI models on Agentverse. Orchestration handled by a central agent sat between user input and other agents, allowing for a modular design that supports future agent integrations. Live data from OpenWeatherMap, Mapbox, and Spotify enriches every response.

Challenges we ran into

Keeping five async agents coherent with each other was harder than expected. We also had to design a full idempotent request lifecycle with per-agent deduplication so partial failures didn't cascade.

Accomplishments that we're proud of

A genuinely usable product demoed under pressure. Progressive card rendering that feels snappy even with live API calls in flight. An anonymous-first architecture that personalizes without requiring a login. A modular design of the agentic workflow that will be able to easily integrate any agent one wishes.

What we learned

From this project, we learned a lot about agentic AI and its capabilities. Working on this project made us realize how big this will be for advancing progress and our engineering capabilities in the future.

What's next for Dayger

Creating a general agent workflow where the user can choose any agent from the marketplace, pull context from our application, and output relevant card information.