Travelary

Inspiration

We all struggle to plan for travel because of our busy schedules and school responsibilities. Thus, we decided to create an AI-powered itinerary to easily plan our transportation, food, and entertainment details without extensive searching.

What it does

Our itinerary generator takes vacation details such as destination, date, travel members, hotel rooms, and goal of travel and creates a custom itinerary that combines your desires with an optimal, time-efficient plan.

How we built it

Gemini Implementation

First, we created our Gemini API key and experimented with a custom AI chat through HTML and CSS. After we were able to get successful input and response between the user and the AI, we began prompting the AI with feedback from the Crawl4AI web scraping.

Crawl4AI Web Scraping

Crawl4AI uses LLM strategies to generate Markdown from sites. We configured it to ignore links and images so Gemini could interpret the data. We created a schema that separated the markdown into formatting components such as:

• History
• Transportation
• Things to Do
• Food
• Travel Routes
• Events and Attractions
• Hotels

The parameters are then inputted into site urls like booking.com and rome2rio.com for Crawl4AI to scrap prices, travel details, and routes. Next, Gemini uses the following details to create short summaries and lists of information.

Website Design

Initially, we used the Simple.css stylesheet to design a basic website with input fields and formatted sections to display our information. We worked on creating a Figma mockup to recreate in HTML, but struggled with the translation.

Challenges we ran into

We originally were using BeautifulSoup, a Python library that extracts data from HTML, but were unable to scrap sufficient data from booking websites. Also, we used Python Selenium, a web driver, to select and input data onto booking websites so we could get more advanced details about locations and travel information. However, most sites limited our ability to scrape the available data.

Accomplishments that we're proud of

We are very proud of our web scraper and combination with Gemini because it provides an elaborate summary of data and a great starting point for planning a vacation, including times, costs, and possible events.

What we learned

We learned how to use software such as Flask and Figma to help create web apps.

What's next for Travelary

We wanted to develop a mobile version of Travelary to be more accessible to users.

Built With

• crawl4ai
• css
• flask
• html
• python