ChowChat

• 

  ChowChat interface

• 

  ChowChat answering a prompt about the chicken under $20 near the user

Inspiration

I'm a huge food lover, but here's my problem with Google Maps, DoorDash, and UberEats: none of them let me just ask "where can I get pad thai?" and get a simple list with prices. Instead, I have to scroll through dozens of restaurants, tap into each one, dig through their entire menu, and manually compare prices. When I'm hungry and craving something specific, I don't want to wade through endless options—I just want to know where I can get my dish and how much it costs.

What it does

ChowChat is a conversational chatbot that searches restaurant menus and prices for you. Just tell it what dish you're craving—like "pad thai" or "chicken tikka masala"—and it instantly returns a curated list of restaurants that serve it, along with their prices. No more endless scrolling or menu hunting. How we built it We started with Vercel AI's one-step deployment for the front-end and chatbot integration. Then we connected the chatbot to our restaurant dataset and configured it to parse menu queries through natural language prompts. The chatbot processes your request, searches through our database, and returns relevant matches with pricing info.

Challenges we ran into

1/ Houston dataset{link](https://www.kaggle.com/datasets/graphquest/restaurant-menu-items ): was clean but too limited – We initially tried using a Houston restaurant dataset, but it was way too small and didn't have variety on locations. So you can't search for "Near me" 2/ UberEats dataset cleanup was brutal – We pivoted to scraping UberEats data, but cleaning and structuring it took way longer than expected and ate into our development time.

Accomplishments that we're proud of

Built a working chatbot that actually solves a real problem we experience constantly Successfully integrated real restaurant data despite dataset challenges Created a simple, intuitive interface that gets you answers in seconds instead of minutes

What we learned

• There are trade offs: the houston dataset with more detailed descriptions enabled more detailed queries but was limited to menus and restaurants in houston while the UberEats datasets were less detailed in their descriptions but enabled a wider search, ultimately supporting the "near me" search functionality.
• Natural language processing for food queries has unique challenges (synonyms, variations, abbreviations)
• Sometimes the simplest user experience requires the most backend complexity

What's next for ChowChat

• Expand to more cities and integrate live menu data from multiple delivery platforms
• Add filters for dietary restrictions, price ranges, and restaurant ratings
• Implement location-based recommendations and delivery time estimates

Built With

• claude
• next.js
• openai
• postgresql
• vercel