All Projects
SeeU Bangkok
AI-powered Bangkok trip discovery with maps, chat, saved itineraries, and multilingual planning.
Overview
SeeU Bangkok is an AI-powered travel discovery platform for finding places, planning trips, and exploring Bangkok through an interactive map and chat experience. It combines a Next.js web app, a Hono API, Supabase, and AI agent workflows to help users discover authentic local spots instead of only the usual tourist highlights.
Problem & Solution
The Problem
Bangkok trip planning is often fragmented across maps, blogs, booking sites, and manual notes. That makes it difficult to answer simple questions about nearby places, budgets, routes, hidden gems, and multilingual support in one workflow.
The Solution
SeeU Bangkok unifies discovery and planning into one flow with an interactive map, conversational AI planning, saved itineraries, Supabase-backed accounts, and streaming responses so users can discover places and build trips in real time.
Tech Stack
Provides the web app shell for place discovery, itinerary planning, saved trips, and the user-facing experience.
Handles the backend API surface for place search, itinerary planning, and streaming AI responses.
Stores user sessions, saved trips, and planner data while supporting authentication and persistence.
Drives the interactive map experience so users can discover places visually and plan around location context.
Keeps the monorepo contracts, API validation, and frontend state management predictable across the stack.
Supports a responsive UI that stays lightweight and fast across desktop and mobile trip-planning flows.
Key Features
- AI chat for natural-language place discovery and itinerary planning
- Interactive map-based exploration with nearby suggestions and category filters
- Saved trips, chat sessions, and editable itineraries for returning users
- Supabase-backed authentication and user-specific planner state
- Streaming AI responses over SSE for real-time planner feedback
- Thai and English support for a multilingual travel experience
- Admin tools for maintaining place data and content
Challenges & Learnings
Balancing map UI responsiveness
Map-heavy screens need to stay responsive while cleaning up Mapbox instances correctly and avoiding jank on desktop and mobile.
Coordinating streaming AI workflows
The planner has to coordinate search, tool calls, and partial responses without breaking UI state or making the chat feel unstable.
Combining context-aware recommendations
Useful travel recommendations need category, proximity, budget, timing, and route structure at the same time instead of simple keyword search.
Supporting multilingual and authenticated usage
The product has to work smoothly for Thai and English users while still handling login, saved trips, and planner persistence cleanly.