SeoulMate: An Intelligent Seoul Travel Itinerary Generation System Based on FastAPI and LangChain

SeoulMate is an intelligent Seoul travel itinerary recommendation backend service based on FastAPI, LangChain, and OpenAI GPT. Its core is to automatically generate personalized travel routes by combining user preferences, real-time crowd data, and high-quality merchants recommended by the Seoul Metropolitan Government, solving the pain points of traditional guides being too general and unable to adjust dynamically.

Modern tourists pursue personalized, localized, and dynamically adjustable itineraries. Traditional guides are too general to adapt to personal preferences, real-time situations, and budgets. The SeoulMate project emerged as a solution: using FastAPI as the framework, integrating LLM capabilities with real-time data, and combining Seoul's official high-quality affordable merchants ("착한가격업소") and real-time attraction crowd density information via LangChain and the OpenAI GPT-4o-mini model to provide tailored route recommendations.

SeoulMate uses a Python backend tech stack, with core components including:

1. FastAPI framework: Provides high-performance asynchronous processing, automatic data validation, and API document generation, supporting Android client requests (forwarded via Spring Boot);
2. LangChain and OpenAI integration: Uses LangChain to achieve structured output, ensuring AI-generated content conforms to predefined JSON formats;
3. Pydantic data models: Defines models like GoodPriceStore and MeetupCongestionDto for data validation and LangChain output schemas;
4. Middleware and logging: Custom trace_id_middleware generates unique trace IDs, and structured logs record key operations and AI interactions.

The itinerary generation logic integrates multi-dimensional data:

• User preference parsing: Receives travel dates, interest tags, number of people, budget, and natural language special needs;
• Data fusion strategy: Integrates Seoul's official high-quality merchant data (priority recommendation), real-time crowd density (to avoid crowded areas), and AI knowledge base (to supplement itinerary content);
• Structured output: Guides GPT via System Prompt to generate responses in the CourseResponse format, including itinerary charm introductions and location lists with coordinates for direct frontend display.

The application value of SeoulMate:

• Tourists: Solves language barriers and information asymmetry; natural language requests can get professional itineraries, and real-time crowd data improves travel comfort;
• Local merchants: Priority recommendation of officially certified high-quality merchants brings precise customer flow and promotes the local economy;
• City tourism management: Demonstrates the combination of government open data and AI technology, enhancing the intelligence level of urban tourism services.

Project technical highlights:

1. Modular design: Separates configuration, logging, middleware, and business logic for easy maintenance and expansion;
2. Environment variable management: Sensitive information (such as OpenAI keys) is managed via .env files, supporting multi-environment configuration;
3. Type safety: Python 3.10+ type annotations and Pydantic models catch type errors during development;
4. Extensible architecture: Reserved extension points allow access to other city data, multi-language support, and integration of real-time weather/traffic information.

SeoulMate is an excellent case of combining traditional web development with cutting-edge AI technology. It demonstrates the practice of building production-level AI applications with FastAPI + LangChain, and the value of integrating government open data with commercial AI. It provides a full-stack reference for developers and shows the potential of AI personalized recommendations for tourism technology practitioners. In the future, it can integrate multi-modal (image recognition) and voice interaction functions to further enhance the user experience.