Kontour.ai - Journey Reimaged For Human and AI.

Main Poster for Kontour.ai
Destination planning demo for a French Riviera trip scenario.
Import an itinerary online into Kontour AI to instantly convert external trip plans into editable day-by-day workflows.
Welcome screen introducing the Kontour AI planning workspace.
AI trip context analysis view with extracted traveler intent.
Extended trip context analysis with refined recommendation signals.
Bot-friendly skills interface that enables AI agents to execute structured travel planning actions reliably.
Add a new itinerary activity with AI-assisted details and scheduling.
Map-based location finder for trip planning and POI discovery.
Hotel discovery and booking options via Sabre-connected workflow.
Live flight search demo powered by Sabre API integration.
End-to-end flight booking flow with full Sabre API results.
Unified export panel for multiple itinerary output formats.
Structured JSON export for itinerary portability and integrations.
Printable PDF itinerary export for offline travel use.
Export itinerary stops directly into Google Maps format.
Google Maps handoff after itinerary export and redirect.
Central sharing panel for text, image, video, and widget outputs.
Share itinerary as clean, copyable text.
Share itinerary as a simple visual image.
Poster-style itinerary card for social sharing.
Generate a video-ready frame for itinerary storytelling.
Exported video frame asset ready for publishing.
Embed itinerary as an interactive web widget.
Export package for deploying the itinerary web widget.

About the Project

Kontour.ai started from a simple frustration: travel planning keeps getting smarter, but also more fragmented. Chat tools talk. Maps visualize. Booking engines transact. None of them think together. Humans are forced to stitch meaning across tabs, screenshots, PDFs, and redirects. That cognitive tax felt unnecessary in an era where AI can already reason across complex systems.

The inspiration was not to build a better travel planner, but to build a travel reasoning workspace. One place where intent, geography, time, and tradeoffs coexist. A place where AI does the heavy orchestration, and humans stay in control of judgment.

What We Built

Kontour.ai is an AI-powered travel planning and reservation workspace that merges chat, map, and itinerary into a single continuous loop.

At its core, the system treats a trip as a living state machine rather than a static plan. Each user input updates context. Each AI response is grounded in session memory, spatial constraints, and itinerary structure. Nothing resets. Nothing disappears.

We implemented:

A psychologist-guided conversational engine that uncovers intent progressively, reducing user fatigue.
Map intelligence that reasons about routes, sequences, and feasibility, inspired by operations research solvers.
A GDS-connected, bot-friendly reservation layer capable of placing real bookings without breaking context.
A structured itinerary model that supports creation, editing, sharing, and multi-format export.
Social sharing flows where users publish evaluations, not just plans, enabling creator-style distribution and earning.

The UI is deliberately calm: chat for intent, map for truth, itinerary for commitment.

How We Built It

The frontend is built with Lit + Vite for performance, clarity, and long-term maintainability. State persistence and migration were treated as first-class concerns to support multi-session reasoning.

Gemini is deeply integrated, not abstracted away. We rely on:

Long-context reasoning for multi-turn planning.
Multimodal understanding for map and itinerary coordination.
Structured output to keep chat clean while powering internal orchestration.

Session management, context extraction, and activity orchestration live in dedicated services, ensuring the system scales without turning into a monolith.

Conceptually, the system follows:

$$ \text{Trip State}_{t+1} = f(\text{User Intent}_t, \text{Context}_t, \text{Spatial Constraints}, \text{AI Reasoning}) $$

The AI proposes. The system validates. The human decides.

What We Learned

We learned that better AI is not about more autonomy, but better boundaries. Users trust systems that explain why, not just what. Keeping humans as evaluators dramatically increases confidence and satisfaction.

We also learned that real-world integration matters. Recommendations feel hollow if they cannot be executed. Booking, state persistence, and sharing are not extras; they are the proof that an AI system understands reality.

Challenges We Faced

The hardest problems were not UI or prompts, but coordination:

Keeping chat outputs expressive while preventing leakage of internal reasoning.
Synchronizing map, chat, and itinerary without race conditions or user confusion.
Designing booking workflows that are both automation-friendly and human-readable.
Maintaining session continuity across refactors without breaking user trust.

Each challenge reinforced the same lesson: AI products fail when they feel scattered. Kontour.ai succeeds by staying whole.

Why This Matters

As AI grows more powerful, humans will spend less time operating tools and more time evaluating experiences. Kontour.ai is built for that future. AI handles complexity. Humans keep taste.

Travel becomes less about planning, and more about living.

More Technically

AI-powered travel planning workspace that combines chat, map, and itinerary management in a single interface. We are proud to use full Google API family, including Gemini, Maps, Flights, etc... The project is built with Lit + Vite and includes session persistence, context extraction, activity orchestration, sharing, and booking-oriented workflows.

Core Systems

1) Chatbot Intelligence

Psychologist-guided, subtle question sequencing to progressively uncover trip intent without overwhelming users
Context-aware recommendation flow designed to improve relevance, planning confidence, and decision quality turn-by-turn

2) Map Intelligence

Smart route calculation and map orchestration integrated into planning flows
AI-generated navigation summaries for clearer travel decisions
Route optimization inspired by operations research solver strategies

3) Reservation (GDS-Connected, bot friendly)

One-stop booking workflow that can place real orders across travel components
Bot-friendly booking interfaces and skill-compatible workflows for seamless automation

4) Itinerary and Sharing

Structured itinerary planning and day-level activity coordination
Creator-to-follower sharing through text, poster, video frame, and embeddable web widget formats