Smart_Itenary_planner

Inspiration

What iInspiration

Planning trips for both individuals and small teams usually means endless tabs, scattered chats, and messy spreadsheets. Smart_Itinerary_Planner was born to fix that: one conversational AI that can plan a quick solo getaway, a friends’ trip, or a startup offsite with the same intelligent backend.

What it does

Smart_Itinerary_Planner is a conversational, AI-powered itinerary engine for individuals and groups. Users talk naturally (“3 friends, 2 days in Goa, thrill + seafood, not too packed”), and the system progressively fills a structured checklist (dates, budget, group type, preferences, constraints) until it has enough signal to generate a realistic, day-wise plan. It returns a complete itinerary with activities, timing, budget breakdown, and map-ready GeoJSON routes that a modern UI can visualize.

How we built it

The backend is a Node.js/Express API that exposes a chat endpoint plus itinerary endpoints. A Groq-powered LLM (Llama 3.3 70B) handles natural language understanding, extracting structured fields into a TripChecklist and deciding the next best question to ask. A conversation engine tracks sessions, completeness, and missing fields, then triggers itinerary generation once core fields are filled. The data model (sessions, POIs, itinerary days/blocks, budget, GeoJSON) is designed to plug into Raindrop SmartBuckets for travel content and SmartMemory for persistent user preferences.

Challenges we ran into

Designing a conversational flow that feels smart, doesn’t repeat questions, and still systematically fills a strict checklist was non-trivial. Getting the model to return clean JSON with both natural questions and machine-readable fields required careful prompt design and robust parsing. Balancing generic travel logic with realistic constraints for Goa/India (budgets, local transport, time windows) also required good defaults and mock data while keeping the system extensible.

Accomplishments that we're proud of

We built a stateful conversational planner that actually “knows” when it has enough information and when to ask something specific next. The checklist-driven design keeps the AI explainable and debuggable, while still feeling natural for users. The backend is cleanly separated from the UI, with testable REST endpoints, a clear itinerary schema, and map-friendly output ready for a polished frontend for both individual and team trips.

What we learned

A strong data model (checklist, sessions, POIs, itinerary blocks) matters as much as the model choice for building reliable AI products. Small prompt tweaks—like explicitly choosing a nextField or defining critical fields for “ready to plan”—dramatically improve UX. We also saw how Raindrop SmartComponents can serve as the missing backend for AI apps: SmartBuckets for curated travel data, SmartMemory for user/session state, and SmartInference to orchestrate multi-step planning.

What's next for Smart_Itinerary_Planner

Next steps include wiring real SmartBuckets with destination-specific POIs, integrating routing (OSRM/GraphHopper) for accurate travel times, and using SmartMemory to remember each traveler’s style across trips. On the UX side, we plan a React-based map view with interactive day timelines, real-time tweaks (“swap this beach”, “lighten day 2”), and shareable links for teams and friend groups. Longer term, Smart_Itinerary_Planner aims to evolve into a general “trip brain” that powers multiple frontends—from web dashboards to voice-first travel guides.t does

Built With

• express.js
• gorq
• raindrop
• react