Milky Way

LOOM VIDEO: link What it does Milky Way generatively connects the vast information from different aspects of your life in an interactive galaxy of expandable nodes. Users can import context from your Spotify jams to your Google browser history to your Gmail inbox, from which we extract your 'north stars' and most meaningful motifs. Milky Way collects and compacts the cosmic dust of your digital footprint so that you can dive deeper into what makes you, you and explore and share your digital life-scape through a playful, interactive interface. We hope to make your universe a little more reflective and a little less lonely. So, grab a friend, hop on a rocket ship, and discover what's in your orbit :)

Inspiration

Introspection and interconnection. Generative AI should serve to expand human creativity and curiosity instead of replacing it. We wanted to build an aesthetic interface for humans to explore the intricate details they may not have realized about themselves, and have the ability to express those details to others.

How we built it

Frontend: React Backend: FastAPI, Composio (for Gmail), Chrome History MCP, Spotify API, OpenAI

Challenges we ran into

Integrating search, Gmail, and Spotify APIs and creating replicatable logic for including more user-provided context in the future. Recursive logic for building the graph representation of the solar system. OAuth flow coordination across multiple providers with different token refresh mechanisms Rate limiting from Spotify and Gmail APIs when processing large historical datasets Parsing and normalizing inconsistent data formats from Chrome history vs. structured API responses

Accomplishments we're proud of

Synthesizing and filtering through various personal contexts to extract meaningful patterns Authentication orchestration across three separate OAuth providers with secure token management Dynamic clustering algorithm that groups related activities into coherent "planets" using embeddings Infinite zoom functionality with lazy-loading and progressive detail revelation

What we learned

Efficient display for graph networks using canvas rendering and spatial indexing Tool calling for data retrieval with proper error handling and fallback strategies Managing async data pipelines while keeping the UI responsive Privacy-first architecture (all analysis happens client-side after initial data retrieval)

What's next

Support for additional data sources (Calendar, Photos, Beli, Snapchat, Strava, LinkedIn) Intrastellar orbit: find your common planets with your friends Universe expansion—visualize how your galaxy has evolved over months/years Generative search––prompt-based searching and planet creation Mobile app with AR mode to view your galaxy in physical space ^^ (oo )_____ (__)\ )\/\ | |---w | | | | |