Inspiration
Make travel planning feel like a conversation — ditch forms and menus and let people plan trips by talking naturally to an AI travel agent.
What it does
Tripmate listens to your voice, extracts destination and dates, transcribes the exchange live, and generates a ready-to-use itinerary that opens on your dashboard.
How we built it
Frontend: React + Vite + TypeScript with Tailwind for UI. Voice & NLU: Deepgram and a voice-agent layer for live transcription and intent extraction. FetchAI's ASI:One llm for drafting the itenary. Backend: Node + Express, PostgreSQL for storage. Real-time flows connect the voice agent to the itinerary generator.
Challenges we ran into
Turning messy, spoken input into reliable structured data (dates, city names) and keeping the UI, agent, and backend perfectly in sync during live conversation.
Accomplishments that we're proud of
A seamless voice-first flow that extracts intent and dates reliably, live split-screen transcription (AI vs. user), and automatic generation + redirection to a polished itinerary.
What we learned
Voice-first UX is powerful; real-time sync is tricky and benefits from clear state boundaries and graceful fallbacks for ambiguous input.
What's next for Tripmate
Add multi-destination and group planning, booking integrations (flights/hotels), deeper personalization (preferences, budgets), calendar/phone sync, and a mobile-first voice experience.
Built With
- asi:one
- deepgram
- javascript
- leaflet.js
- node.js
- postgresql
- python
- react


Log in or sign up for Devpost to join the conversation.