Image of Chrome Extension on Travel Webpage
Image of Chrome Extention when collecting locations from webpage

Itinera - Manahil and Avani

Inspiration

As avid travelers and planners, we’re all too familiar with the struggles of juggling several windows with hundreds of tabs, all while going back and forth between spreadsheets and websites collecting destinations of interest. The monotonous copy-pasting process and grouping of locations was tedious and took hours and days of research leading to the ideation behind Itinera.

What it does

Itinera is a Chrome extension that scrapes the websites the User opens once they click ‘START’, to collect content from the web page. Through the extension’s plugin code, the text data is sent into the OpenAI API, where the plugin prompts ChatGPT to extract specific locations and attractions mentioned. These results are returned to the User. The User selects their locations of interest through the extension and may select ‘SAVE’ which stores all locations in a database, allowing them to apply the extension on multiple tabs at once. When they are satisfied with their selected locations across all tabs, they may click “GENERATE” and enter the duration of their stay. Then they will be redirected to view their itinerary on ChatGPT where they can work directly with the GenAI to edit their itinerary. No further web page data is collected by the extension.

How we built it

We began by breaking down the overall project into manageable pieces we could start hacking away at - Phase 1, the scraping of the webpage by the Chrome Extension and connecting GenAI platforms to extract locations, and Phase 2, the importing of the locations and preferences into the GenAI platform to generate an itinerary. For Phase 1, we began by coding the component of the Chrome Extension that reads from the current webpage and saves the text. This required exploring various methods of creating an efficient, well-packaged Chrome Extension in JavaScript and HTML and understanding the ways the extension interacts with a web page. With the extension successfully extracting all the text, we explored how Chrome Extensions could connect to ChatGPT for Phase 1. This involved finetuning our prompts to ChatGPT to efficiently list locations in a streamlined manner by working with the Chrome extension code to submit the prompts. In the meantime, we worked on Phase 2 of the project, taking the collected locations and building a personalized itinerary based on the duration of their stay. This required further prompt fine-tuning as well. We continued to research how to weave all elements of the project together to form a single streamline process from opening a web page and clicking “START” to building the final itineraries with the power of GenAI.

Challenges we ran into

As early developers ourselves, we attended our first Hackathon with hopes of creating technology we would be passionate about - both to develop and to use. With our idea in mind, we set out exploring and researching all the necessary technologies and connections, while also focusing on making the user experience as smooth and as intuitive as possible. Understanding and experimenting with these technologies was a clear learning curve, where every step of the way required new documentation to read up, more tutorials to watch, and new technology to play with. Further, navigating the intricate world of OpenAI and various APIs was a daunting task, alongside the pressure of building such a complex project ground up.

Accomplishments that we're proud of

Yet, beyond these challenges, this experience has provided us with the opportunity to move beyond our comfort zone and put all our effort into making a solution that we are highly passionate about. We are proud that, even without a fully deployed solution now, we were able to make strides in increasing our technical knowledge and have a deeply fleshed-out plan to bring this solution successfully to reality soon. We are also deeply satisfied with our work together as a team, and how we could combine our skills and put them to the test when building our project. We are excited to bring this project to reality and are highly proud of the progress we have made so far.

What we learned

As first-time hackers, we learned the effort it takes to create a proof of concept, taking the user experience into account. We learned with every challenge we ran into, using our passion for the concept to carry us into late-night progress checks and demos.

What's next for Itinera - Travel AI

We have several potential features we’d like to implement in Itinera, all for greater convenience, applications, and improvement. For example, we aim to incorporate Google’s MyMaps feature into the extension for quick, organized visualization of itineraries. Further, we will develop this technology to easily collect destinations of interest from liked/favorited posts on social media such as Instagram. On the existing extension, we will add the ability to track the sentiment of an article’s opinions on each location and assign a score to each destination accordingly. Further refinements include the ability to save/reload and revisit previously saved locations and to allow various levels of planning, from simply grouping locations through planning to the minute.