Travel Buddy

App Logo
Architecture Diagram
Multi Agent Tool
Travel Buddy App

Inspiration

One of our team members recently came back from a self-planned overseas trip. She found it a hassle to plan everything from scratch, including locating nearby popular restaurants both on social media and Google Maps, grouping the to-go locations in different areas of the city. She was also caught surprised when it rained, forcing her to reshuffle her itinerary, and eventually missed out some spots that she wanted to visit. After the trip, she felt that the use of AI Agents could help in crafting a comprehensive travel plan that will make every penny spent on the trip worthwhile.

What it does

Travel Buddy is a one-stop solution that crafts an end-to-end personalized travel itinerary. If you have a place to visit in mind, you'll just have to tell Travel Buddy, and it will create an extensive plan for you - from which flight to take, to what time to enter a restaurant. Fret not about the weather, nor the distance to the next travel destination. Every detail of the travel itinerary will be meticulously taken care of, and you'll just have to follow through the instructions.

How we built it

The team, comprising of three individuals having zero exposure to AI Agents, started off by watching 3 hours tutorial on ADK. Then, we divided and conquered multiple AI agents that we envision to have in Travel Buddy. Finally, we merged them all under a single manager.

How we deploy

The project is deployed using two separate containers on Google Cloud Run. One for the Streamlit frontend and another for the Travel Buddy backend. Each component is built into its own Docker image and deployed independently.

The Streamlit app handles the user interface and communicates with the Travel Buddy backend over HTTP by calling its public Cloud Run URL. This containerized, serverless setup ensures scalability, clean separation of concerns and easy deployment while enabling seamless interaction between the frontend and backend services.

Challenges we ran into

One of our biggest challenges was understanding how to structure multiple agents and get them to work together cohesively. It took time to figure out how to break down the travel planning process into modular agents that could handle different tasks like flight search, attraction grouping, restaurant discovery, and weather-based itinerary reshuffling.

Another hurdle was data integration. Pulling useful and structured information from different sources (e.g., weather APIs, restaurant listings, and flight options) and making the agents reason effectively over them introduced complexities. Handling real-world variables such as unpredictable weather or closed attractions added further layers to the agent logic.

Lastly, debugging multi-agent workflows was tricky. Since each agent operated semi-independently, tracking how data flowed between them and identifying which step caused errors required careful design and extensive testing.

Accomplishments that we're proud of

With the use of Travel Buddy, the struggle of planning a trip, especially for any solo travel newbie, can be shortened from months to minutes. The highlight of Travel Buddy is the integration of various factors that are often overlooked when planning an itinerary, including weather forecast and automated grouping of travel destinations in the same area.

Now that the team member has come back from the trip, she realize she should have developed Travel Buddy with the team much earlier ago! The itinerary Travel Buddy provides is similar to her actual itinerary.

What we learned

With the completion of this project, we have gained much technical knowledge about AI Agents, especially the use of ADK. We particularly appreciate the ease to test AI Agents using adk web, which sped up the project progress significantly.

What's next for Travel Buddy

We’re excited to continue developing Travel Buddy beyond the hackathon version. Here are a few directions we're exploring:

Live Data Integration: Hooking into real-time APIs for flight booking, weather, and transportation to make the itinerary more responsive and up-to-date.
User Preferences and Profiles: Allowing users to define dietary preferences, travel styles (e.g., budget, luxury, adventure), and pace (relaxed vs. packed), which will let Travel Buddy craft even more personalized itineraries.
Mobile Experience: Bringing Travel Buddy to a mobile app so users can access, modify, and follow their itinerary on the go.
Collaborative Planning: Supporting travel planning for groups where preferences and availability need to be aligned across multiple travelers.