Finn - Vacation Planner ✈️

Extension for Finn AI agent which recommends users trips

Comment

Inspiration

We were inspired by the frustration many people face when planning vacations—too many options, hidden costs, and generic recommendations that don’t consider personal lifestyle or financial reality. We wanted to create something smarter and more personalized. With bunq’s rich financial data and the power of LLMs, we saw an opportunity to create an AI agent that could take the stress out of trip planning.

What it does

Finn – Vacation Planner ✈️ is an AI travel assistant that crafts personalized vacation plans based on your:

  • bunq transaction history
  • budget constraints
  • travel preferences
  • current mood

It engages users in a natural chat interface, analyzes their financial behavior, and recommends curated trips—whether you're craving a beach holiday, a cultural city break, or a spontaneous getaway. The assistant supports dynamic conversation, allowing you to tweak trip details, add companions, and more—all through intuitive dialogue.

How we built it

  • Frontend: Built using Angular and SCSS for a clean, responsive, and modular UI on desktop and mobile.
  • Backend: Developed with FastAPI in Python, integrating with the bunq Python SDK (Sandbox) to simulate real user financial data.
  • LLM Integration: We used the meta/llama-3.3-70b-instruct model via NVIDIA’s API to power natural conversations and generate vacation suggestions tailored to each user.
  • We also maintain chat history and detect follow-up modifications, ensuring contextual consistency in multi-turn conversations.

Challenges we ran into

  • Maintaining Chat Context: Ensuring that Finn remembers and modifies travel plans based on follow-up prompts was tricky. We implemented a mechanism to track and adjust previous responses intelligently.
  • bunq API Authentication: Manually signing requests and working in Sandbox mode required careful handling of encryption and session tokens.
  • Connection between frontend and backend - the frontend was designed to resemble a typical bunq interface with an AI agent

Accomplishments that we're proud of

  • Creating a fully functional AI agent that personalizes trips using real (sandboxed) financial behavior.
  • Seamlessly integrating LLM, bunq API, and custom business logic to deliver a rich user experience.
  • Designing an elegant and intuitive UI that works smoothly across devices.
  • Implementing chat-based contextual memory, enabling true back-and-forth vacation planning.

What we learned

  • How to build AI-based applications with real financial APIs like bunq and integrate them meaningfully.
  • The importance of maintaining dialogue context and memory in LLM-driven apps to improve user experience.
  • Real-world challenges in combining financial data with natural language understanding to offer relevant and responsible suggestions.
  • How to work fast, iterate quickly, and turn an ambitious idea into a working prototype in just 24 hours.

What's next for Finn - Vacation Planner ✈️

  • 🪪 Bunq Login Integration: Let users authenticate with their real Bunq accounts and pull real-time data.
  • ✈️ Booking Integration: Enable instant flight, hotel, and activity booking—prioritizing eco-friendly and Bunq-affiliated options.
  • 📱 In-app Integration: Bring Finn directly into the Bunq mobile app as a built-in travel planner.
  • 🌐 Multilingual Support: Make Finn accessible for users across Europe and beyond.
  • 🧠 More AI Features: Integrate smarter suggestions based on real-time trends, weather, and global events.
  • 📅 Calendar Sync: Automatically add trips to users' calendars and send reminders and packing lists.
Share this project:

Updates