mapwise

Inspiration

When you're out walking, biking, or exploring a new area, you often find yourself wondering: "What's that building?" or "Is there something interesting nearby?" Mapwise was born from that everyday curiosity — a desire to effortlessly access real-time information about the world around you without having to stop and type.

What it does

Mapwise uses your real-time location and voice input to answer natural language queries about your surroundings. By combining OpenAI’s real-time APIs with geospatial data, it can identify landmarks, describe nearby places, and even surface relevant local news — all with a simple spoken question.

How we built it

We used Next.js for the frontend, integrated OpenAI’s real-time APIs for transcription and response generation, and tapped into Tavily to provide local news and context-aware content. It’s a location-aware, voice-first experience designed to feel effortless and conversational.

Challenges we ran into

Getting voice input to reliably convert to text across different browsers

Accomplishments that we're proud of

Successfully integrating Windsurf for real-time voice interactions
Building a working, voice-controlled conversational map assistant
Creating a fluid UI without relying on Google Maps navigation

What we learned

How to make a live conversational agent feel natural and responsive
How to handle (and recover from) AI hallucinations and errors
The power and quirks of Windsurf, Lovable, and OpenAI’s real-time APIs
That building a multimodal experience is very possible — and very fun

What's next for mapwise

Supporting longer conversations with context memory
Letting users save their conversations and revisit their voice history to find points of interest
Enhancing map personalization and proactive discovery suggestions