Inspiration

Booking a home service shouldn't require downloading 5 apps, filling out forms, and making phone calls. We wanted to make it as simple as sending a text message. What if you could just say "My AC is broken" and an AI agent handles everything — finding vendors, comparing ratings, and booking — all in one conversation?

What it does

FixIt is an AI-powered home service booking agent. Users describe their problem in plain English, and the AI:

  • Classifies the issue into one of 6 categories (AC Repair, Plumbing, Electrician, House Cleaning, Pest Control, Appliance Repair)
  • Searches and ranks top-rated vendors by location, price, and reviews
  • Books appointments instantly through the chat
  • Provides real-time booking confirmation and tracking
  • Handles in-chat payments

How we built it

  • Frontend: Next.js 16 with React and TypeScript, featuring a custom glassmorphic design system (Lumina Frost) with pastel gradients, frosted glass panels, and Material Symbols icons
  • Backend: Python with FastAPI serving the AI agent pipeline
  • AI Engine: Google Gemini with function calling for natural language understanding, intent classification, and multi-step tool orchestration
  • Design: Stitch MCP for UI prototyping, then hand-integrated into production React components

Challenges we ran into

Orchestrating multi-step AI tool calls (search vendors → compare → book → confirm) reliably was the biggest challenge. We also had to ensure the glassmorphic UI stayed performant with heavy backdrop-filter blur effects across all Android screen sizes.

What we learned

Function calling with LLMs is incredibly powerful for building conversational agents that take real-world actions. We also learned that a well-defined design system makes frontend development 10x faster.

What's next for FixIt

  • Real-time technician tracking with live map
  • Voice input support
  • Multi-language support (Hindi, Spanish)
  • Integration with real vendor APIs and payment gateways

Built With

Share this project:

Updates