Inspiration
My brother is on the spectrum, and so are several of my close friends. I've watched them do the same exhausting routine before every dinner out: scrolling through dozens of Yelp reviews, searching for keywords like "quiet" or "not too crowded," trying to piece together whether a place will be overwhelming or okay.
Even when reviews say a restaurant is "cozy," that doesn't tell you if the music is blasting or if the tables are crammed together. One bad experience (walking into a place with harsh lighting and a 45-minute wait in a packed lobby) can ruin the whole night.
I built Comfrt because the people I love deserve better than guesswork. They deserve to know before they arrive whether a place will actually feel comfortable for them.
What it does
Comfrt combines the Yelp AI API with Claude AI to create a hybrid AI system for finding sensory-friendly venues. Here's how it works:
- Search naturally - Ask questions like "Find a quiet coffee shop to work from" or "Calm Italian restaurant for a date"
- Conversational refinement - Follow up with "anything cheaper?" or "what about closer to downtown?" without repeating yourself
- Get comfort scores - Each venue gets a 0-100 comfort score with AI-generated insights unique to that place
Key Features:
- Hybrid AI Analysis: Yelp AI handles natural language search, Claude analyzes reviews for sensory comfort signals
- Multi-turn Conversations: Refine your search naturally without starting over
- Noise Level Predictions: Know if it's quiet, moderate, or loud before you arrive
- Comfort Map View: Color-coded map showing calm (green) to lively (red) venues
- Best Time to Visit: Predictions for when venues are calmest
- Outing Planner: Chain multiple comfort-friendly stops with a total comfort score
- Shareable Cards: Generate beautiful cards to share your favorite calm spots
How we built it
The frontend is built with Next.js 15 and React 19, styled with Tailwind CSS 4 for a warm, calming aesthetic.
The Hybrid AI Stack:
- Yelp AI API (
/v2/ai/chat) powers the conversational search experience. Users ask natural language questions and get intelligent responses with business entities. Full multi-turn context means follow-up questions just work. - Claude AI (Anthropic API) analyzes the returned venues and their reviews for sensory comfort signals that Yelp doesn't explicitly track, like when reviewers mention "we could actually hear each other" or "the lighting was soft and relaxing"
Supporting Infrastructure:
- Yelp Fusion API for additional business details, attributes, and reviews
- Mapbox GL for the interactive comfort map
- Supabase for analytics tracking
- Vercel for deployment
The comfort scoring algorithm weighs multiple factors: Yelp's noise_level attribute, ambiance categories, Claude's review sentiment analysis, venue type base scores, and user preferences.
Challenges we ran into
The biggest challenge was making two AI systems work together seamlessly. Yelp AI returns great venue recommendations, but we needed Claude to analyze the why behind comfort. We had to carefully prompt Claude to look for indirect signals. People rarely say "sensory-friendly." They say things like "perfect for a quiet date" or "not too hectic even on weekends."
Another challenge was maintaining conversation context. Yelp AI handles multi-turn conversations beautifully, but we had to ensure our comfort scoring stayed consistent across follow-up queries.
Accomplishments that we're proud of
- The hybrid AI approach genuinely surfaces insights neither system could provide alone
- Multi-turn conversations feel natural. "Anything with outdoor seating?" just works.
- My girlfriend actually uses it now before we go out. That's the best validation I could ask for.
What we learned
Combining multiple AI systems creates something greater than the sum of its parts. Yelp AI excels at understanding search intent and venue recommendations. Claude excels at extracting nuanced comfort signals from unstructured review text. Together, they solve a problem neither could solve alone.
What's next for Comfrt
- Community reviews: Let users rate venues specifically on sensory comfort
- Real-time crowd data: Integrate with live data for current conditions
- Mobile app: Native iOS/Android for on-the-go searches
- Shareable Saved favorites: Personal lists of verified calm spots
Log in or sign up for Devpost to join the conversation.