Your personalized travel guide is here! Travel-guide uses four Hogwarts House agents to handle it in minutes

[Mitchellhk]

This third prize-winning work of the Multi-Agent Hackathon, travel-guide-network, can generate travel guides for you in no time!

Travel-guide-network is a multi-agent travel guide system built on the OpenAgents framework.

As long as you input the city you want to visit and travel dates, the four agents corresponding to Hogwarts College will each provide travel advice such as scenic spot recommendations, outfit choices, and risk control from their respective perspectives.

Simply put, it’s like hiring four butlers with different personalities to give you advice at the same time.

The project details are as follows, come and take a look!

About the Author
A current independent developer and a long-term technology explorer. My project is a travel guide system based on weather information, which can generate diverse travel suggestions in the Huo Gewoci style according to the weather. I hope it can become a simple travel companion, making travel more interesting.

Project Description
This project builds a multi-agent travel guide system based on the OpenAgents framework.

The core idea is to use YAML configuration to define different role instructions, enabling multiple agents to concurrently process the same real-time weather data and generate differentiated personalized recommendations.

The system is orchestrated in a lightweight manner through Python scripts, combined with the Ollama local model and a strict source_id verification mechanism, achieving a loosely coupled collaboration model and stable data flow control, allowing complex workflow construction without writing low-level business logic.

Technical Solution
The system is orchestrated by combining Python with YAML configuration. The data entry point is based on WorkerAgent to build the Weather Connector HTTP service, with requests triggered by the travel_sender script. Model invocation uses the Ollama local service.

In terms of data flow, the Connector publishes weather data in JSON format, which is concurrently received by multiple downstream Agents. These Agents strictly filter message sources through the source_id verification mechanism to ensure they only respond to specific instructions, effectively preventing the infinite loop issue in a multi-Agent environment.

Functional Features
Real-time weather data scraping: Automatically obtain weather information based on the input city and date.

Automated instruction trigger: Supports one-click initialization of network layer and Agent processes and sending requests via command-line scripts.

Multi-agent concurrent processing: Supports five or more Agent nodes to process the same input data simultaneously.

Differentiated Recommendation Generation: Based on different System Instructions (such as different college styles), output customized guides that include dressing suggestions, activities, and precautions.

System Stability Assurance: Through the source ID verification mechanism, avoid message chaos and loops in a multi-Agent environment.

Github:
https://github.com/starttown/travel-guide-network