MicroItinerary — AI Travel Planner

Problem

Travel apps optimize for proximity and ratings. They don't consider human energy levels, group dynamics, or intelligent budget allocation.

Why It Matters

Trip planning is tedious and error-prone. Most apps just list options — they don't understand your constraints. This uses AI to actually solve the planning problem.

System Approach

• Spring Boot 3.2.2 + Java 21 backend with Flyway migrations
• React 18 + Vite PWA frontend with offline support via IndexedDB
• OpenAI GPT-4 integration for intelligent destination and cost suggestions
• Redis caching to reduce API costs and improve response times
• Google OAuth 2.0 + JWT for secure authentication

Key Decisions & Trade-offs

• PWA over native app — broader reach, single codebase, works offline
• OpenAI API adds latency and cost, but rule-based alternatives lack quality
• Expense splitting algorithm prioritizes simplicity over Splitwise feature-parity
• PostgreSQL 16 for relational data with potential for vector search later

Current Status

Backend API functional. Frontend PWA with offline sync working. AI suggestions integrated.

Roadmap

• Add calendar view for annual planning
• Implement real-time expense tracking
• Add collaborative trip editing

What I'd Improve Next

• Could add caching layer for repeated AI queries
• Consider fine-tuning a smaller model for cost estimation