Travel Planner: Your Smart Travel Designer

QWEN CHAT

DISCORD

Introduction#

We are excited to introduce our brand-new Travel Planning Assistant, a powerful system built on a Multi-Agent architecture with robust real-world tool-calling capabilities. It is designed to handle complex and dynamic itinerary planning tasks with high efficiency. Whether you’re preparing for a multi-city trip or an in-depth single-city exploration, it can deliver precise, actionable travel plans:

• Daily schedules down to the hour: covering transportation, accommodation, attractions, dining, and other key aspects.
• Total budget estimates: plan your expenses in advance.
• Important travel reminders: such as weather alerts, attraction regulations, and transportation updates.
• Verified real-world data: all routes and information are traceable and reliable.

Key Features#

🧠 Multi-Agent Architecture: Efficient Planning for Complex Trips#

Travel planning is a systematic task involving transportation, attractions, lodging, dining, and more — each interconnected and mutually influential.
A single agent cannot easily capture this complexity.

To address this, we built a sophisticated Multi-Agent collaboration framework that efficiently manages dynamic interactions among modules.
With intelligent coordination, the system adapts seamlessly to diverse travel scenarios:

• Single-city deep tours (e.g., Beijing)
• Specialized trips at large scenic areas (e.g., Qiandao Lake)
• Multi-city routes (e.g., the Xinjiang circuit)

🔍 Authenticity: Every Piece of Information is Traceable#

The Travel Planning Assistant integrates real-world third-party tools (such as Amap, Fliggy and web search) to provide:

• Accurate flight, train, hotel, and attraction information.
• Real-time traffic updates, attraction opening status, and ticket prices.
• Precise calculations of distances and travel durations.
• Nearby restaurant recommendations.

During the generation of a 3-day itinerary, the system may call tools more than 70 times, ensuring authenticity and reliability in every plan.

🔁 Outstanding Planning and Reflection Capabilities#

Powered by the strong reasoning and self-reflection abilities of the Qwen model, the system not only creates accurate query plans proactively at the start but also continuously validates and optimizes itineraries during execution.

Planning Phase:

• Predefines query targets according to the type of trip:
  • Transportation methods
  • Attraction opening hours
  • Routes and distances between sites
  • Meal insertion points

Reflection Phase:

• Continuously refines the plan based on real-time data from sub-agents:
  • Late-night or next-day train arrivals → Automatically adjust attraction order to maximize efficiency.
  • Attractions closed on Mondays → Reschedule to another day and optimize the overall route.

Demo#

Demo 1: Qingdao five-day trip#

Demo 2: Xi'an five-day trip#