Background Story

I have recently participated in the 5-Day Gen AI Intensive Course organized by Google and Kaggle. I was taught about prompting, Vector Databases and RAG (Retrieval Augmented Generation) Schematic JSON Responses, Chat History Storing, function calling and many more AI concepts. Now it is time to apply these and try to create a Capstone Project.

Building WanderWise

In the era of personalized experiences, what if planning your dream trip was as easy as chatting with a friend? That’s exactly what WanderWise, our AI-powered travel assistant, aims to do.

In this blog post, we'll walk through how we built WanderWise using Google's Gemini API, Python, and Streamlit—all from the comfort of a Kaggle notebook. Whether you're a hobbyist or a product builder, this guide offers everything you need to create and deploy an intelligent itinerary planner.

Functional Flow

Tools and Technologies Used

• Google Generative AI (Gemini API)

• Streamlit for UI

• Google Colab for development

Concepts used from the Course

• Structured output/JSON mode/controlled generation

• Few-shot prompting

• Agents

• Function Calling

Understanding the Workflow

We actually have to Flows - End to End Flow and the only Itinerary Flow.

End To End Flow we can Include all things that a Traveler needs in his Trip Planning like Flight, Train or Bus tickets, Local Commute Cabs, Hotels etc. These can be integrated via APIs from tech Giants like Make My Trip, Amadeus, Sabre and other players like Airbnb, Sky Scanner etc.

Second the only Itinerary flow is where we only help user create the Itinerary for his trip and then it is upto him to choose the mode of transportation and stay options.

This whole idea can be integrated as a chat-bot to your Website like Make My trip, Amadeus SECO, ClearTrip etc.

User Preference Gathering

We try to ask The user for some initial 10-12 questions to understand his basic expectation from his trip and then suggest the Initial Itinerary

Initial Questions Asked are

1. Who are you traveling with?

2. What kind of weather do you prefer?

3. What kind of experience are you looking for?

4. What budget range are you considering?

5. Preferred accommodation style?

6. How packed do you want your days to be?

7. Any specific cuisines you want to try?

8. Any dietary restrictions?

9. Do you prefer places where English is widely spoken?

10. Any other preferences or requirements you'd like to share?

The responses of these questions is stored in a JSON object which is used to create a initial Itinerary and then we have the chat window open in which the user can chat and make conversational prompts to the bot to edit the itinerary and when the itinerary is finalized the chat window is closed with a final draft.

Chat-bot Images and Video

Future Advancements

1. We can take Stay Prefrences and suggest best stays like Hotels, Airbnbs, Hostels

2. You can also add Cab and Package Services

3. Encourage User to book Flights, Trains or Bus from your Platform

4. Sell Whole Itiernary as a Package