built in a one-week hackathon
RepairDispatch
2026
An AI coordinator for service requests, from first message to booked visit.
For teams that coordinate repairs. A request comes in by chat or voice, coverage is checked, a contractor found and the visit booked, and anything the rules can't settle goes to a human operator.
screens
key decisions
ai on a leash
The LLM only handles the conversation. Everything that costs money is a rule that can be tested and explained.
strict output
AI output must match a schema and pass validators before anything is saved. A bad answer costs an operator a minute, not a booking.
one write path
Every change goes through one function that also writes the audit entry.
where ai comes in
ai in the product
- ›Conversational intake with structured output
- ›Voice input with speech-to-text
ai in the build
- ›Built mostly through an AI app builder, steered by an agent rulebook
stack
tanstack startreacttypescriptsupabaserealtimedrizzlecloudflarellm apispeech-to-text