Plotted

• 

  Chat with an AI to plan your trip — and watch it come alive on the map.

Inspiration

Planning a trip is exciting, but it's also exhausting. You're juggling browser tabs, Reddit threads, Google Maps, and budget spreadsheets all at once. We wanted to collapse that entire process into a single conversation. Just tell the AI where you want to go — it asks a few smart questions, builds a day-by-day itinerary, and pins every location on an interactive map in real time.

What it does

TravelMap AI is a conversational travel planner. You type a destination, and the AI:
Asks personalized questions — days, budget, travel style — before generating anything
Builds a structured day-by-day itinerary with times, durations, and notes for each stop
Pins every location on a live Mapbox map, with routes and transport info between stops
Surfaces real Reddit tips from city-specific subreddits for each place
Exports your full itinerary as a Word document (.docx) — ready to share or print
Supports follow-up conversations — say "make it 3 days instead" and it replans on the spot

How we built it

Next.js 16 (App Router) — full-stack framework
Qwen Plus (via OpenAI-compatible API) — LLM powering the chat and itinerary generation
Mapbox GL / react-map-gl — interactive map with custom markers and route lines
Reddit API — surfacing real community tips for each destination
docx + file-saver — client-side Word document export
Tailwind CSS v4 — styling

Challenges we ran into

Getting the LLM to reliably output structured JSON itineraries — required careful prompt engineering and a normalization layer to handle malformed responses Keeping the chat feel natural while maintaining multi-turn state for itinerary follow-ups Mapbox geocoding for obscure POIs — some place names needed fallback search strategies

Accomplishments that we're proud of

The end-to-end flow feels genuinely useful: one conversation → map + exportable doc The conversational itinerary refinement (e.g. "actually I want 3 days") works seamlessly Reddit integration adds real, human-sourced context that generic AI answers miss

What we learned

Structured output from LLMs requires defensive parsing, not just trust Map UX and chat UX are surprisingly complementary — seeing your plan on a map makes it feel real

What's next for Plotted

Support for more destinations with live LLM planning Flight and hotel search integration Collaborative trip planning (share a link, edit together) Mobile app

Built With

• mapbox
• next.js
• react
• reddit
• tailwindcss
• typescript