SYYL - 2emojis - Beacons

Turn strangers into neighbors. Anywhere, anytime

Comment

Inspiration

Our app was inspired by the simple yet powerful idea of bringing people closer together in their everyday surroundings. By removing barriers and making it effortless to connect with those nearby, we create shared, meaningful experiences that foster community and connection. Our AI goes beyond general assistance by focusing specifically on location-based questions and interactions, helping people navigate, discover, and make their daily experiences easier and more informed in any given place.

What it does

Our app helps people connect effortlessly by scanning a QR code to join location-based chat rooms, breaking down barriers to meeting those around us. Whether in classrooms, libraries, or parks, it acts as an icebreaker, helping start conversations digitally before leading to real-life interactions. Users can ask location-specific questions by tagging their messages. The AI uses context from past questions and answers about that location, accessed through a retrieval-augmented generation system with a vector database. This means it doesn’t learn on its own but pulls relevant information based on the most popular, community-validated responses. The community also participates by sharing their knowledge, making the answers more accurate and helpful. To keep conversations welcoming, every message is moderated for inappropriate content, fostering a safe space. Through these shared experiences, the app builds stronger connections and a more interconnected community where neighbors, classmates, and strangers become friends.

How we built it

  • We have 3 repos, one each for the backend, frontend, and the content moderation azure function
  • We used Azure cosmos DB to store chats
  • azure SignalR servicefor a realtime chat experience
  • Azure content moderation service + azure function to moderate chats for inappropriate messages
  • Azure app service and vercel for deployment
  • Vertex AI RAG Engine to store context and make it easy to search on it ## Challenges we ran into
  • Making AI retain context in an efficient way to help with responses for location based messages
  • Content moderation a bit annoying
  • brainstorming and coming up with an innovative idea

Accomplishments that we're proud of

  • Ai responses to location tagged questions, where the AI learns from peoples responses to provide more accurate information (uses most upvoted responses for confidence)
  • Theres is content moderation on every chat people send (inappropriate stuff will be deleted)
  • The real time text experience looks quite cool :) ## What we learned
  • we learned a ton about nitty gritty details about Azure portal and GCP
  • Learned about new cloud services that were useful for our project ## What's next for Beacons
  • Make it so that only people in that location can be in chat rooms
  • fix some refreshes, make better UI
  • Mobile apps for android and IOS
  • Keep improving the AI context based on peoples Data
  • potentially create our own AI models once we have enough data on peoples responses for Location tagges messages
  • Redis cache for faster in memory operations
  • easier ways for people to create and manage rooms like teachers creating rooms with QR code for the class
  • improve chatting experience, @ people and allow people to post images(moderated with Azure AI Content Safety)

Ai was used to help, debug and code parts of the project

Built With

  • anthropic
  • application-insights
  • asp.net-core-8
  • auth.js-(nextauth)
  • azure-app-service
  • azure-content-safety
  • azure-cosmos-db
  • azure-signalr-service
  • c#
  • dotnetenv
  • github-actions
  • google-custom-search
  • google-oauth
  • google-recaptcha-v3
  • html5-qrcode
  • javascript
  • next.js
  • node.js
  • postcss
  • qrcode
  • react
  • react-markdown
  • remark-gfm
  • signalr
  • swashbuckle-(swagger)
  • tailwind-css
  • typescript
  • typescript-tooling
Share this project:

Updates