SearchTour is an AI-powered travel assistant platform designed to address core challenges in the tour-based travel industry. Tour operators need to create, manage, and market complex multi-day travel products, including itineraries, pricing, departures, policies, images, and SEO content, while also handling bookings and distribution. At the same time, travelers need a more flexible way to explore and evaluate tours, often asking multi-part questions about group composition, pricing, travel styles, and conditions.
The project solves both sides within a single system. It supports multi-day tours (2–30 days), private and group travel, customizable itineraries, and structured product management, including creating, editing, and organizing tour data. The traveler-facing AI assistant enables natural, multi-step queries and returns grounded answers based on real tour, policy, and destination data. In parallel, an operator-facing assistant supports content generation and product authoring workflows, improving how tours are structured, described, and prepared for discovery and sales.
Technically, the system is a full-stack software application built with Next.js (TypeScript) and FastAPI. It uses a Retrieval-Augmented Generation (RAG) architecture with ChromaDB, combined with deterministic backend tools and LangGraph-based agentic workflow orchestration. The assistant follows a controlled execution model inspired by principles from “Building Effective Agents,” where the LLM is used for planning, routing, and composition, while repository content and backend logic remain the source of truth.
This project demonstrates how AI assistants and agentic systems can be applied to real-world travel platforms, where accuracy, structure, and system design are critical. Next steps include booking workflows, personalization, and pricing intelligence.
